Brigham Young UniversityBYU ScholarsArchive

All Theses and Dissertations

2016-06-01

Analyses of mRNA Cleavage by RelE and the Roleof tRNA Methyltransferase TrmD Using BacterialRibosome ProfilingJae Yeon HwangBrigham Young University

Follow this and additional works at: https://scholarsarchive.byu.edu/etd

Part of the Chemistry Commons

This Dissertation is brought to you for free and open access by BYU ScholarsArchive. It has been accepted for inclusion in All Theses and Dissertationsby an authorized administrator of BYU ScholarsArchive. For more information, please contact scholarsarchive@byu.edu, ellen_amatangelo@byu.edu.

BYU ScholarsArchive CitationHwang, Jae Yeon, "Analyses of mRNA Cleavage by RelE and the Role of tRNA Methyltransferase TrmD Using Bacterial RibosomeProfiling" (2016). All Theses and Dissertations. 6443.https://scholarsarchive.byu.edu/etd/6443

Analyses of mRNA Cleavage by RelE and the Role of tRNA Methyltransferase TrmD

Using Bacterial Ribosome Profiling

Jae Yeon Hwang

A dissertation submitted to the faculty of

Brigham Young University

in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

Allen R. Buskirk, Chair

Barry M. Willardson

William R. McCleary

Joel S. Griffitts

Joshua L. Andersen

Department of Chemistry and Biochemistry

Brigham Young University

June 2016

Copyright © 2016 Jae Yeon Hwang

All Rights Reserved

ABSTRACT

Analyses of mRNA Cleavage by RelE and the Role of tRNA Methyltransferase TrmD

Using Bacterial Ribosome Profiling

Jae Yeon Hwang

Department of Chemistry and Biochemistry, BYU

Doctor of Philosophy

Protein synthesis is a fundamental and ultimate process in living cells. Cells possess

sophisticated machineries and continuously carry out complex processes. Monitoring protein

synthesis in living cells not only inform us about the mechanism of translation but also deepen

our insights about all aspects of life. Understanding the structure and mechanism of the ribosome

and its associated factors helped us enlarge our knowledge on protein synthesis.

Recently, with the dramatic advances of high-throughput sequencing and bioinformatics,

a new technique called ribosome profiling emerged. By retrieving mRNA fragments protected by

translating ribosomes, ribosome profiling reveals global ribosome occupancy along mRNAs in

living cells, which can inform us with the identity and quantity of proteins being made. Easily

adapted to other organisms, ribosome profiling technique is expanding its application in

revealing various cellular activities as well as the knowledge on protein synthesis.

Here, we report the mechanism of translating mRNA cleavage by endoribonuclease RelE

in vivo. RelE is an endoribonuclease that is induced during nutrient deficiency stress and

specifically cleaves translating mRNAs upon binding to the ribosomal A site. Overexpression of

RelE in living cells causes growth arrest by inhibiting global translation. We monitored RelE

activity in vivo upon overexpression using ribosome profiling. The data show that RelE actively

cuts translating mRNAs whenever the ribosomal A site is accessible, resulting in truncated

mRNAs. RelE causes the ribosome complexes to accumulate near the 5’ end of genes as the

process of ribosome rescue, translation, and cleavage by RelE repeats. RelE cleavage specific

sub-codon level ribosome profiling data also represent reading frame in Escherichia coli and

sequence specificity of RelE cleavage in vivo.

We report another ribosome profiling study on a methyltransferase TrmD in E. coli.

TrmD is known to methylate G37 (the residue at 3’ side of anticodon) of some tRNAs and be

responsible for codon-anticodon interaction. We constructed a TrmD depletion E. coli strain,

whose deletion results in lethality of cells. Resulting depletion of m1G37 in the strain leads to

growth arrest. Lack of m1G37 of some tRNAs whose codons start with C showed frequent

frameshift when translating the gene message in vitro. By using ribosome profiling, we

successfully observed significant difference on translation process when codons interact with

anticodons of tRNAs lacking m1G37. The data reveal slow translation rate or pauses on the

tRNAs when missing the appropriate methylation, which corresponds to the previous

biochemical data in vitro.

Keywords: RelE, endoribonuclease, TrmD, methyltransferase, ribosome profiling, m1G37

ACKNOWLEDGEMENTS

I had opportunities to experience three different laboratories across the U.S, which is not

common for a graduate student. After working two years at BYU, I worked at Scott Strobel’s

laboratory at Yale for about 10 months, then back to BYU for a year, and again to Rachel

Green’s laboratory at Johns Hopkins for two years so far. Moving from one place to another with

my family slowed down the progress of my research but, with the help from many people I could

continue studying.

I would like to first thank my former lab members at BYU. I still remember when I talked

to Chris Woolstenhulme for the first time while searching labs to join. He helped me to have an

idea what our lab was doing for research and helped me to learn many lab skills later on. Mickey

Miller was always nice and helped me to understand better what I was doing in the lab. He was

always patient in answering many questions I had.

I express thank you to the lab members at Yale University, who helped me to understand

kinetic assays for RelE activity and shared their findings from experiments. I am grateful for

their encouragement for me while I was struggling with continuous failing on my experiments.

They helped me to realized that I was not the only one who was discouraged with repeated

failures by sharing their own experiences. I also appreciate their talents and creativities in

approaching problems and carrying experiments.

I would like to thank current lab members at Johns Hopkins for the discussions and for

sharing their talents. With their mature citizenship and efficient setup at the lab, I had an

excellent environment to focus on the research. I would like to also express thank you to Isao

Masuda for the collaboration on TrmD study and for his friendship during the work at Ya-Ming

Hou’s lab at Thomas Jefferson University. I also thank Ya-Ming Hou for sharing her knowledge

on tRNAs and methylations.

I am grateful for the generosity of Scott Strobel and Rachel Green. They both allowed me

to work at their labs. I am honored to be part of their labs, being inspired by their insights and

superb attitudes towards science in proximity. Their enthusiasm and knowledge on the field

stimulate me to have passion on science as well. Rachel always talks about science without

ceasing and her energy and confidence remind me of what I want be.

I also thank for my committee members at BYU for their guidance. I especially

appreciate Barry Willardson for his attitude towards science and for teaching ethics in science

through a class. He told a group of students in a class, including me, that science is built on the

foundations of previous researchers so that we are obligated to share what we obtain. That lesson

was one of the most valuable lessons for me during graduate school years. I thank Joel Griffitts

for demonstrating his critical thinking in solving problems. I am also thankful for BYU and

funding sources that supported me to study during the program. I have been impressed by many

great people at BYU for their teachings and care for the students.

Most of all, I deeply thank my advisor and mentor, Allen Buskirk, who provided me with

many opportunities to learn. He taught me that science is a really fun thing that I should be

excited about. I like his sense of humor and the way he put things together in science. His talent

and knowledge on science make him special and that, somewhat, separates me from him, but I

always aimed to overcome the gap between mine and his and to catch up the level of his

proficiency, which still looks a long way. He has guided and helped me on the course with great

patience. I feel like I am just ready to start real science now.

Lastly, I would like to thank my family and friends for their support. I would like to

thank Kyu Tae Han and Keum Hee Song, who helped me broaden my views and shape my mind.

I extend special thank you to my eldest brother, Jang Yeon for his support since I was young.

Without his help, it would not be possible to come to the U.S. and continue to study what I

always wanted to. I am so grateful for my parents for their constant love and trust towards me. I

love my family, my beautiful wife, Jee Yeon Lee, my daughters, Yiwon and Jiwon, who always

give me motivation and strength to continue. There were hard times during the course and they

were always with me to encourage me. Without the love and trust from my wife, I would not be

able to accomplish this and will not be able to go further.

vi

TABLE OF CONTENTS

List of Tables ............................................................................................................................... viii

List of Figures ................................................................................................................................ ix

1 Introduction .................................................................................................................................. 1

1.1 Bacterial Translation .............................................................................................................. 1

1.1.1 Initiation ......................................................................................................................... 2

1.1.2 Elongation Cycle ............................................................................................................ 3

1.1.3 Termination .................................................................................................................... 4

1.1.4 Ribosome Recycling ....................................................................................................... 5

1.1.5 Ribosome Rescue ........................................................................................................... 5

1.2 Ribosome Profiling ................................................................................................................ 8

1.2.1 The Ribosome Profiling Procedure ................................................................................ 9

1.2.2 Analysis of Ribosome Profiling Data ........................................................................... 10

1.2.3 Power of Ribosome Profiling to Study Protein Synthesis ............................................ 11

1.2.4 Bacterial Ribosome Profiling: Promises and Obstacles ............................................... 14

1.3 The Endoribonuclease RelE ................................................................................................. 17

1.4 tRNA Methylations and TrmD ............................................................................................ 19

2 mRNA Cleavage by RelE........................................................................................................... 20

2.1 Introduction .......................................................................................................................... 20

2.2 Results and Discussion ........................................................................................................ 23

2.2.1 Ribosomes are Enriched at the 5’-end of Genes in Cells Overexpressing RelE .......... 23

2.2.2 RelE Predominantly Cleaves mRNA after the Second Nucleotide in Empty A sites .. 28

2.2.3 in vitro Digestion with RelE Reveals the Reading Frame ............................................ 31

2.2.4 RelE Prefers to Cleave after C and before G ................................................................ 34

2.2.5 The Specificity of RelE Interferes with Analyses of Ribosome Pausing ..................... 37

2.2.6 Refined RelE-Derived Ribosome Density Better Reflects Reading Frame ................. 39

2.2.7 Concluding Thoughts ................................................................................................... 41

vii

2.3 Materials and Methods ......................................................................................................... 42

2.3.1 Bacterial Strains and Growth Conditions ..................................................................... 42

2.3.2 Preparation of Ribo-seq and RNAseq Libraries ........................................................... 42

2.3.3 Data Analysis ............................................................................................................... 43

3 The Role of tRNA Methyltransferase TrmD ............................................................................. 44

3.1 Introduction .......................................................................................................................... 44

3.1.1 tRNA Synthesis and Post-Transcriptional Modifications ............................................ 45

3.2 The TrmD Methyltransferase ............................................................................................... 47

3.3 Results 49

3.3.1 Achieving Regulation of TrmD, an Essential Protein .................................................. 49

3.3.2 Depletion of m1G37 of tRNA Causes Pauses on CCG, CCA, and CGG Codons........ 53

3.4 Materials and Methods ......................................................................................................... 57

3.4.1 Strains and Construction .............................................................................................. 57

3.4.2 Cell Growth and Preparation of Ribo-seq Libraries ..................................................... 58

3.4.3 Data Analysis ............................................................................................................... 58

References ..................................................................................................................................... 59

viii

LIST OF TABLES

Table 1 Codons and their anti-codons with modifications ........................................................... 56

ix

LIST OF FIGURES

Figure 1 Average ribosome occupancy......................................................................................... 24

Figure 2 Ribosome enrichment at 5’-end genes of operon ........................................................... 26

Figure 3 Average ribosome density at start- and stop-codon ....................................................... 28

Figure 4 Ribo-seq data at nascent peptide-mediated stalling sites ............................................... 30

Figure 5 Average ribosome occupancy for a wild-type sample ................................................... 32

Figure 6 Ribo-seq and RNAseq density at each position ............................................................. 33

Figure 7 Sequence bias at the 3’-end of ribosome footprints ....................................................... 35

Figure 8 Sequence bias at the 3’-end of RNA fragments ............................................................. 36

Figure 9 The ribosome occupancy on codons............................................................................... 37

Figure 10 Correction for better reading frame .............................................................................. 39

Figure 11 Programmed frameshifting at the prfB gene ................................................................ 40

Figure 12 1-methyl-Guanosine ..................................................................................................... 47

Figure 13 TrmD depletion strain construction .............................................................................. 50

Figure 14 TrmD depletion ............................................................................................................ 51

Figure 15 Growth measurements upon TrmD depletion .............................................................. 52

Figure 16 Effect of TrmD depletion on tRNA .............................................................................. 53

Figure 17 Ribosome occupancy on codons in control strain ........................................................ 54

Figure 18 Ribosome occupancy on codons in degron strain ........................................................ 55

1

1 INTRODUCTION

1.1 Bacterial Translation

Ribosomes read the genetic information in mRNA and facilitate peptide-bond formation

to make proteins from single amino acids. Escherichia coli ribosomes are composed of two

subunits termed 50S and 30S because of their sedimentation coefficients. The larger 50S subunit

is itself composed of two ribosomal RNAs (23S and 5S) and 34 ribosomal proteins; the active

site that links amino acids together is found within the large subunit. The smaller 30S subunit

contains the 16S ribosomal RNA (rRNA) and 21 proteins and plays the dominant role in

decoding the genetic information in mRNA. The two subunits combine to make 70S ribosome

complexes that are active in translation.

Peptide synthesis occurs in the ribosome in a very sophisticated way with the help of key

adaptor molecules known as transfer RNAs (tRNAs). tRNAs are first charged with activated

amino acids to form aminoacyl-tRNAs (aa-tRNAs). There are 20 different aminoacyl-tRNA

synthetases (aaRSs) in E. coli1. Each aaRS activates the correct amino acid with ATP and then

links the amino acid to its cognate tRNA by an ester bond between the carboxyl group of the

amino acid and the 3’-terminal adenosine of the tRNA. The aa-tRNAs then enter the ribosome,

and when bound to their cognate codons, deliver the appropriate amino acid for the ribosome to

incorporate into the growing polypeptide chain. The ribosome has three sites for tRNA binding:

the aminoacyl-tRNA or A site, where the correct aa-tRNA is selected, the peptidyl-tRNA or P

2

site, where the growing nascent chain is attached, and the exit or E site, through which

uncharged tRNAs pass as they are released from the ribosome. The protein synthesis process is

complex and continuous but it can be conveniently broken down into four discrete steps:

initiation, elongation, termination and ribosome recycling2.

1.1.1 Initiation

At least in most cases, free 50S and 30S ribosomal subunits assemble into 70S complexes

at start codons to initiate protein synthesis. Initiation factor 1 (IF1) binds to the A site of the 30S

subunit and promotes dissociation of the 30S and 50S subunits. Initiation factor 3 (IF3) binds to

the E site of the 30S subunit and further prevents premature association with the 50S subunit3.

The 30S subunit bound with IFs interacts with the purine-rich region known as the Shine-

Dalgarno (SD) sequence on mRNA that is complementary to the 3’-end of the 16S rRNA4. The

pyrimidine-rich region of 16S rRNA, or the anti-Shine-Dalgarno (aSD) sequence, binds to the

SD sequence roughly 7 nt upstream from the start codon5. This interaction positions the start

codon of the mRNA at the P site of the 30S subunit. In bacteria, start codons are predominantly

AUG although GUG is also sometimes used. The fact that ribosomes can be recruited to many

internal sites within a polycistronic mRNA differentiates bacterial translation from eukaryotic

translation, where ribosomes are recruited by proteins that bind the 5’-cap and then scan and

initiate at the first start codon in mRNAs that are, as a rule, monocistronic6.

Guanosine 5’-triphosphate (GTP)-bound initiation factor 2 (IF2) recognizes the initiator

tRNA, N-formyl-methionyl-tRNAfMet (fMet-tRNAfMet)3, 7. IF2 binds in the A site of the 30S

subunit and places the initiator tRNA in the P site of the 30S subunit. Binding of initiator tRNA

to the start codon at the P site causes conformational changes of the 30S subunit that lead to

release of IF3. With the release of IF3, the 50S subunit can bind to the 30S subunit, inducing

3

GTP hydrolysis by IF2. At that point, IF1 and GDP-bound IF2 are released from the 30S subunit,

leaving behind an active 70S initiation complex. With initiator tRNA at its P site and an empty A

site, the complex is ready for elongation.

1.1.2 Elongation Cycle

At the start of the elongation cycle, the 70S initiation complex with initiator tRNA bound

to its P site accepts the next aa-tRNA into its empty A site8. aa-tRNAs are delivered to the

ribosome by the abundant GTPase, elongation factor-Tu (EF-Tu)9. Upon GTP binding to EF-Tu,

EF-Tu arranges its three domains in a compact structure allowing them to bind aa-tRNA10. When

the ternary complex (EF-Tu, GTP, and aa-tRNA) binds to the A site, it is rejected if the

anticodon of the tRNA does not match the codon in the A site8, 11. This process repeats until the

cognate tRNA comes in8, 11-12. Upon binding of the cognate tRNA in the A site, the correct

codon-anticodon base pairing causes conformational changes in the ribosome8. rRNA residues

surrounding the base pairs stabilize tRNA binding to the A site8, 12.

Conformational changes in the A site upon binding of the correct aa-tRNA result in a

distortion of the tRNA, triggering GTP hydrolysis by EF-Tu10-11. As the GTP of EF-Tu is

hydrolyzed and the γ-phosphate is released, the structure of the EF-Tu is rearranged and releases

the tRNA, and EF-Tu dissociates from the A site10, 13.

GTP hydrolysis occurs faster with cognate tRNAs than with near- or non-cognate tRNAs;

this kinetic difference is one basis for selectivity for the correct tRNA14. Following GTP

hydrolysis, the aminoacyl end of the A-site tRNA swings into the peptidyl-transferase site of the

50S subunit, a movement known as accommodation15. Cognate tRNAs are accommodated more

quickly than near- or non-cognate tRNAs, as second kinetic selection or proofreading step14-15.

4

RNA nucleotides in the peptidyl-transferase center bring the two amino acids together to

form a peptide bond15. There are two RNA elements in the 50S subunit called the A loop and P

loop that have highly conserved sequences and interact with the 3’-ends of aa-tRNAs positioned

at A site and P site15. Once the C-terminus of growing peptide and the amino group of the A-site

bound aa-tRNA are positioned correctly, peptide bond formation occurs intrinsically by a

nucleophilic displacement, perhaps involving side chains on the tRNAs themselves but not the

rRNA nucleotides in the active site2, 15.

For translation to continue, the ribosome must move to the next codon on the mRNA.

Following peptide-bond formation, the deacylated tRNA is at the ‘P/E’ hybrid state and the

peptidyl-tRNA with two amino acids is at the ‘A/P’ hybrid state15-16. The tRNAs need to be

resolved to a new classical state with the deacylated tRNA completely in the E site and the

peptidyl-tRNA positioned at the P site15. This process is called translocation and is catalyzed by

elongation factor G (EF-G), a ribosome-activated GTPase. Hydrolysis of GTP by EF-G helps the

30S subunit move to the right position for the next codon at A site, maintaining reading frame by

moving exactly three nucleotides on mRNA17. The uncharged tRNA at the E site is released from

the ribosome and another aa-tRNA enters the A site2, 15. This cycle repeats until a stop codon

(UAG, UAA, or UGA) appears at the A site. Translocation is thought to be relatively fast

compared to decoding, and bacterial peptides are formed in the ribosome at a rate of about 20

amino acid residues per second2, 15.

1.1.3 Termination

When the ribosome reaches a stop codon, protein synthesis is terminated. Stop codons are

recognized by two release factors, RF1 and RF218. By mimicking tRNA structurally, these

factors bind in the A site and interact directly with mRNA19. RF1 recognizes the UAA and UAG

5

stop codons and RF2 recognizes the UAA and UGA stop codons20. Release factors hydrolyze the

ester bond on peptidyl-tRNA and release the newly synthesized peptide chain from the ribosome,

leaving uncharged tRNA in the P site2, 20. Both RF1 and RF2 contain a highly conserved

tripeptide motif (GGQ motif) that is responsible for the catalysis of peptide release from the

peptidyl-tRNA at P site20-21. Following hydrolysis of peptidyl tRNA, RF3, a GTPase found in

some but not all bacteria, catalyzes the release of RF1 or RF2 at the end of the termination

process. GTP hydrolysis then promotes the dissociation of RF320-22.

1.1.4 Ribosome Recycling

After termination, ribosomes need to be dissociated for efficient initiation at another start

codon to begin. Ribosome recycling factor (RRF) and EF-G dissociate the ribosome into

subunits, releasing mRNA and tRNAs from the ribosome upon GTP hydrolysis2. The structure of

RRF mimics tRNA in the A site2, 23. When the ribosome dissociates, IF1 replaces the deacylated

tRNA and the mRNA is released. All translational components are now free for another round of

translation23.

1.1.5 Ribosome Rescue

In bacteria, the translation and transcription of genes are coupled. Since genomic DNA is

not separated from the cytoplasm by a nuclear membrane like it is in eukaryotic cells, the whole

process of gene expression happens simultaneously and is faster than it is in eukaryotic cells. As

soon as an mRNA begins to be synthesized, translation can begin. This can lead to problems

when ribosomes translate messages that are defective as when RNA polymerase prematurely

terminates prior to reaching the stop codon in the message. Other defective RNAs come from

6

mutations, chemical damage, and the endonucleolytic cleavage that is part of normal mRNA

decay pathways24.

Truncated mRNAs are deleterious to the cell for at least two reasons: they cause

ribosomes to stall at the 3’-end of the message and produce incomplete or premature proteins

which might be harmful to the cell. When ribosomes reach the 3’-end of a truncated mRNA, RFs

cannot be recruited, since there is no stop codon, and the ribosome-mRNA complex is stably

maintained25. In the cell, stalled ribosomes can affect the global protein synthesis if not rescued.

Bacterial cells have three mechanisms that can resolve stalled ribosomes. One involves a

small, stable RNA known as tmRNA and the other two involve proteins that are alternative

rescue factors, ArfA and ArfB25. tmRNA rescues the stalled ribosome by acting both as a tRNA

and an mRNA25-26. tmRNA-mediated ribosome rescue not only dissociates the stalled ribosome

for recycling but also tags the incomplete nascent peptide for degradation. ArfA is a small

protein that can bind to the A site and recruit RF2, hydrolyzing the nascent peptide and leading

to recycling27. ArfB has a GGQ domain in its structure as RFs do and facilitates the termination

of protein synthesis. Among these three mechanisms, tmRNA is the best characterized28.

The sophisticated structure of tmRNA makes it act as both tRNA and mRNA. In E. coli,

tmRNA is 363 nt long. It contains an tRNA-like domain (TLD) and an mRNA-like domain with

its own small open reading frame (ORF)29. The TLD contains 3’-terminal CCA residues like

tRNAs do and has a G3-U357 wobble base pair, which alanyl-tRNA synthetase recognizes and

charges with alanine at its 3’-end. With several helices and pseudoknots in its structure, tmRNA

is relatively stable against degradation by nucleases. Its partner protein, SmpB, further stabilizes

tmRNA29-30. The structure of the TLD allows EF-Tu to bind Ala-tmRNA just like other tRNAs

do. The GTP-bound EF-Tu-tmRNA-SmpB complex enters the A site of stalled ribosomes and

7

undergoes peptidyl transfer following GTP hydrolysis by EF-Tu30. After the growing peptide is

attached to tmRNA, the mRNA-like domain of tmRNA serves as a transcript for the ribosome to

continue translation. The ORF encodes the ANDENYALAA peptide and a stop codon at the

end31. Because the ribosome switches templates, ribosome rescue by tmRNA is referred to as

trans-translation32. The 11 amino acids added by tmRNA target the aborted nascent peptide for

degradation. The ClpXP protease system recognizes the hydrophobic C-terminal residues of the

tagged peptide and degrades it31.

While tmRNA and SmpB are found in all bacterial genomes, the ArfA backup system

found in some bacteria also facilitate ribosome rescue. It is proposed that ArfA, previously

known as YhdL, binds to the ribosomal A site and recruits RF2. Then, RF2 hydrolyzes the

peptidyl-tRNA and releases translation complex. It is not known, however, how ArfA recognizes

the 3’-end of non-stop mRNA and how it recruits RF2. The deletion of tmRNA and ArfA

together are lethal whereas a single deletion of either is viable, which suggests a mutual role of

the two mechanisms in ribosome rescue26.

ArfB, previously known as YaeJ, also rescues stalled ribosomes. In E. coli ArfB is a

small basic protein with 140 amino acids. As a homolog of RFs, ArfB contains the conserved

GGQ motif for hydrolysis of peptidyl-tRNA28. Instead of having a specific sequence for

interacting with stop codons, ArfB has a positively charged C-terminal tail that interacts with the

mRNA channel, selectively binding ribosomes with little or no mRNA downstream of the A-site

codon. This is reminiscent of the C-terminal tail of the SmpB protein that provides selectivity for

the tmRNA rescue system28.

8

1.2 Ribosome Profiling

Insight into protein synthesis can shed light on many aspects of biology; in particular, it is

a powerful way to monitor gene expression levels in vivo. Previously, researchers have achieved

the goal by monitoring transcriptional changes with microarrays or RNAseq. Another approach,

proteomics based on mass spectrometry (MS), reveals the identity and quantity of proteins in the

cell. However, it is difficult to accurately quantify proteins with low abundance or instability. In

addition, MS focuses on the steady-state levels of proteins in the cell rather than the rate of

protein synthesis. While steady-state levels are very informative in determining the level of

protein activity, they are less helpful in understanding the mechanism of protein synthesis or the

translational control of gene expression33.

Polysome profiling is another method used to monitor protein synthesis in vivo34. This

technique is based on the fact that the number of translating ribosomes bound to mRNAs can tell

us the translational status of that transcript. Ribosome-bound mRNAs are collected by

ultracentrifugation of cell lysates over sucrose gradients. The identity of heavily translated

mRNAs is determined and compared to the background of total mRNA using microarrays or

RNAseq34. Polysome profiling allowed researchers to move from monitoring transcription levels

to monitoring translation levels34. However, the approach has some difficulties in separating

messages depending on the number of ribosome bound to them. Sucrose gradient fraction

collection does not resolve clearly complexes heavier than disomes. The many fractions per

sample (based on the number of ribosomes bound to mRNAs) makes it a lot of work to do the

following analyses. More importantly, the data generated from polysome profiling cannot give

the exact position of ribosomes on mRNAs; rather, they only give some idea of ribosome

abundance on certain mRNAs.

9

To get better information on translation in vivo, Ingolia and Weissman developed a

powerful new technique called ribosome profiling or Ribo-seq35. Simply stated, ribosome

profiling improved polysome profiling by isolating short fragments of mRNA corresponding to

ribosome footprints, revealing the exact positions of ribosomes on mRNAs. Ribosome profiling

provides a genome-wide snapshot of ribosome occupancy on mRNA templates, telling us which

proteins are being made and to what extent35.

The development of ribosome profiling was made possible by advances in sequencing

technology. Ribosome profiling involves deep sequencing millions of ribosome-protected short

mRNA fragments. With the completion of the human genome project, significant improvements

were made in high-throughput DNA sequencing, making the sequencing of hundreds of millions

of short reads economically viable. Methods for preparing cDNA libraries have been well

developed, including RNA ligation methods and techniques for multiplexing a number of

libraries for more efficient and cost effective sample preparation. Another improvement came

from easily accessible databases from bioinformatics. Annotated whole genomes provide the

standard against which sequencing reads can be mapped or aligned. In addition, bioinformatics

has provided many algorithms for processing and analysis of the data generated from high-

throughput sequencing. Together, these technologies and methods make possible the ribosome

profiling technique that is increasing our understanding of gene regulation and particularly the

protein synthesis process.

1.2.1 The Ribosome Profiling Procedure

To collect ribosome-protected mRNA fragments for deep sequencing, grown cells are

effectively captured by flash freezing using liquid nitrogen. For lysis, cells are pulverized

mechanically in liquid nitrogen in the presence of lysis buffer containing drugs that arrest

10

ribosomes on mRNA complexes. The pulverized lysate is thawed and centrifuged to pellet cell

debris. The lysate is then subject to digestion using a nuclease to degrade naked mRNA, leaving

only ribosome-protected mRNA fragments. Ribosome complexes are isolated on sucrose

gradients by ultracentrifugation; digestion with the nuclease has reduced the polysomes to single

monosomes. Ribosome-protected mRNA fragments are released from the monosomes by

treatment with phenol and chloroform followed by RNA precipitation. At this point, the purified

RNAs contains rRNA and tRNA fragments as well as mRNA fragments.

To purify ribosome footprints away from other RNA fragments, the total RNA is run on a

denaturing acrylamide gel for size selection. RNA corresponding to the size that best represents

the footprint of the ribosome are collected. Generally, we isolate mRNA fragments on the gel

that are between 10 to 40 nt in length. The strategy to sequence these mRNA fragments is to

make a cDNA library. First, a defined RNA linker that contains a primer binding sequence is

attached at the 3’-end of each mRNA fragment by ligation with T4 RNA ligase. Linker-ligated

fragments are then gel-purified. rRNA fragments are selectively subtracted by hybridization with

oligonucleotides complementary to rRNA. Then, the purified mRNA fragments are reverse

transcribed to yield cDNAs and these single-stranded cDNAs are circularized by ligation,

amplified by a few cycles of PCR, and submitted for 50 bp, single end, Illumina sequencing.

1.2.2 Analysis of Ribosome Profiling Data

Data analysis involves processing the raw sequencing data into genome-wide maps of

ribosome occupancy and then further analysis for specific needs. High-throughput sequencing is

performed on several samples per lane in a single experiment (a strategy known as

‘multiplexing’). Each library is synthesized with a characteristic sequence (usually 6 nt) termed a

‘barcode’ that distinguishes that library from the others during data analysis. In the first step of

11

data processing, raw sequences are de-multiplexed by sorting them out according to their

barcode sequences. In addition, reads are filtered based on the quality of data as quantified in

Phred quality scores, which is a measure of the quality of the identification of the nucleobases,

and ambiguous or suspicious reads are excluded. After filtering, the reads are trimmed to remove

the constant 3’-linker sequence, leaving only the ribosome footprints. Next, any reads that map

to tRNA and rRNA genes are discarded before mapping reads to annotated genomic DNA. Final

ribosome occupancy data are represented as densities along the genome sequence; these maps of

ribosome occupancy are used for additional analysis using our own computing scripts, which are

written mainly using a computer programming language Python.

1.2.3 Power of Ribosome Profiling to Study Protein Synthesis

Although ribosome profiling has mainly been used to monitor gene expression under two

or more conditions, using ribosome occupancy to report on protein levels, the method has

tremendous potential to shed light on the mechanism of protein synthesis in vivo. There have

been many insights obtained using ribosome profiling since the method was first introduced by

Ingolia and Weissman in 2009. In that first report, several millions of sequencing reads of

ribosome-protected mRNAs were obtained in yeast Saccharomyces cerevisiae. 28-nt long reads

yielded the best information about the position of the ribosome, suggesting that the nuclease

digested completely to the 5’- and 3’-boundaries of the ribosome35. Sequencing reads that

aligned to ORFs revealed ribosome positions on mRNAs with nearly single-nucleotide precision

and showed a strong 3-nt periodicity, providing a means of monitoring the ribosome’s reading

frame35. Genome-wide measurements of ribosome occupancy also generated estimates of the

levels of protein synthesis, which were found to be better predictors of protein abundance by MS

than were measurements of mRNA levels.

12

In a later paper by Ingolia and Weissman, the ribosome profiling technique was adapted

to mammalian cells36. Using the drug harringtonine, which specifically inhibits elongating

ribosomes near the start codon (but not at later sites in the ORF), new translation was inhibited in

cells36. Following the addition of harringtonine, ribosomes were allowed to run off for various

time periods to generate a series of snapshots and reveal a moving picture of translation in vivo36.

Ribosome footprints successfully revealed a progressive depletion of ribosomes from the 5’- to

the 3’-end along mRNAs as the run-off time increased after harringtonine treatment36. The data

showed that ribosomes translate at a rate of 5.6 amino acids per second, consistent with previous

biochemical observations.

In addition to their work on measuring rates of translation with harringtonine, Ingolia and

Weissman also detected sites where ribosome occupancy was strongly enriched due to pausing

during elongation. Programmed nascent chain-mediated pauses were observed on the Sec61b and

Xbp1 genes36. More generally, the PPE/D motif was shown to be associated with internal pauses,

consistent with the slow translation of polyproline stretches. Finally, they also found that

ribosome footprints revealed many alternate ORFs, upstream of canonical ORFs on the same

mRNA, many non-canonical translational start sites (especially at the CUG codon), and evidence

for translation of putative long, non-coding RNAs (lncRNA)36.

High-precision ribosome footprints obtained by ribosome profiling make it possible to

detect programmed frameshift sites. Coupled with a computational method for detecting

transitions between reading frames, ribosome profiling data obtained from human cells revealed

where the same genomic segment is translated in more than one reading frame37. The 3-nt

periodicity of ribosome occupancy shifts if the ribosome has moved into a different reading

frame during translation. Ribosome profiling data is a powerful tool for searching for alternative

13

coding regions in endogenous genes and perhaps in the future in viruses and other mobile genetic

elements rich in such phenomena.

Ribosome profiling in Drosophila melanogaster expanded our understanding of ribosome

readthrough, in which the ribosome continues to translate past a stop codon. Previous studies

have shown that this can occur in a regulated fashion. However, it was very difficult to detect

readthrough; only a small fraction of ribosomes keeps translating, extending proteins past their

predicted C-termini. The presence of ribosome occupancy after stop codons successfully showed

that readthrough is far more than previously predicted in D. melanogaster38.

Ribosome profiling helped to clarify the enzymatic activity of Dom34, a yeast protein

essential for releasing stalled ribosomes39. Dom34 is a homolog of eukaryotic release factor 1

(eRF1) though it lacks a canonical GGQ motif required for peptide release as well as the codon

recognition motif required for interacting with stop codons. Given these differences and its role

in rescuing stalled ribosomes, Guydosh and Green used ribosome profiling to identify specific

targets of Dom34, comparing wild type and dom34 knock-out strains of yeast39. Stalled

ribosomes are stabilized by the lack of Dom34; there are higher levels of ribosome occupancy on

certain mRNAs at stalling sites in the dom34 knock-out strain. The fragments protected by

stalled ribosomes tend to be shorter (15-18 nt) than normal reads. This is because Dom34

preferentially rescues stalled ribosome at the end of truncated mRNAs, meaning that the mRNA

bound to the ribosome is roughly 10 nt shorter at the 3’-end. Furthermore, they found that

Dom34 rescues ribosomes that are found in 3’ untranslated regions (UTRs) of mRNAs39. The

codon-independent rescue mechanism of Dom34 nicely supports the previously known

biochemistry of Dom34, which lacks conserved codon recognition motif39.

14

Ribosome profiling has become a powerful way to understand complex cellular processes

in living cells including protein synthesis. By monitoring protein synthesis in real time, ribosome

profiling has clarified previous biochemical observations and given ample insights on many

aspects. Since 2009, the technique is being more broadly adapted to study various organisms and

the data analysis methods are improving immensely.

1.2.4 Bacterial Ribosome Profiling: Promises and Obstacles

Ribosome profiling has also employed to study bacterial translation. The relatively small

genome of bacteria and less complex gene regulation would suggest that it should be equally

powerful in bacteria as in yeast, where it was first developed. However, there have been several

hurdles to its application in bacteria. The first is that RNase I, the nuclease used to digest naked

mRNA in the yeast protocol, is inhibited by E. coli ribosomes, preventing its use. RNase I

exhibits very little sequence specificity in cleavage and degrades RNA cleanly to the boundaries

of the ribosome, meaning that the 28 nt fragments observed in yeast yield single-nucleotide

resolution and information about the reading frame of the ribosome40. In contrast, all published

bacterial ribosome profiling studies use micrococcal nuclease (MNase), an enzyme with

pronounced sequence specificity40-41. The distribution of fragments generated by MNase is very

broad, with most reads roughly 15 – 35 nt in length, and no 3 nt periodicity or reading frame is

evident. Rather than assigning ribosome occupancy to the 5’-end of reads as was done originally

in yeast, Li and Weissman used a center-assignment strategy in which partial occupancy was

distributed over the length of the entire read, further blurring the position of the ribosome on the

message41. The use of MNase in bacterial ribosome profiling has interfered with an accurate

determination of the ribosomes position, interfering with analyses of ribosome pausing. In spite

15

of these limitations, there have been many advances in understanding bacterial translation using

ribosome profiling, and there have been on-going improvements on the technique itself.

The first ribosome profiling paper using E. coli was reported in 2011 by the Weissman

and Bukau groups in a study of the interaction of the nascent peptide chaperone trigger factor

(TF)40. The authors selectively purified ribosomes bound to TF by cross-linking the nascent

peptide to TF to stabilize transient interactions followed by immunoprecipitation with anti-TF

antibodies. Ribosome footprints from TF-bound and non-TF-bound ribosomes were then

obtained following the regular ribosome profiling procedure40. It turns out that trigger factor

binds about 120 codons downstream from the translational start site, contradicting previous

models that TF interacts as soon as the peptide comes out from the exit tunnel of the 50S subunit

in vitro. Selective ribosome profiling in E. coli by cross-linking the chaperone TF with the newly

synthesized polypeptide revealed valuable information of the kinetics of protein synthesis in

vivo.

Another early ribosome profiling study in E. coli claimed that the main source of pausing

during elongation in bacteria is Shine-Dalgarno-like sequences within open reading frames41. SD

motifs accounted for 70% of strong pauses in their dataset—sites where ribosome occupancy

were enriched more than 10-fold over the gene average. The authors concluded that elongation is

retarded by transient base pairing between SD motifs within ORFs and the aSD sequence in 16S

rRNA41. However, Mohammad et al. contradicted the finding with improved footprint

resolution42. The precision of the ribosome position is significantly improved by assigning

ribosome occupancy using the 3’-end of sequencing reads (3’-assignment) rather than

distributing occupancy over the whole read (center-assignment)42. This simple adjustment seems

to better position the ribosome occupancy along its mRNA transcript on all of known E. coli

16

footprint data so far, lining up the peaks seen at start and stop codons as well as programmed

pauses like the one observed in the SecM protein. It seems that MNase does a better job in

cleaving mRNAs up to the boundary of the ribosome at the 3’-end site of the bound mRNA,

resulting in more tight cleavage at the 3’ end but relaxed cleavage at the 5’ end of ribosome-

protected mRNAs42.

With the higher resolution obtained by 3’-end assignment it became clear that the signal

observed by Li and Weissman could be split into two separate signals due to two distinct

phenomena: first, ribosomes pause with Gly codons in the E site, perhaps due to the nature of the

inhibitor in the lysis buffer used to arrest translation. Second, the bona fide SD pauses arise from

preferential enrichment of long RNA fragments from the population of ribosome footprints. It

turns out that mRNAs that interact with the aSD in 16S rRNA are protected from digestion at the

5’-end and so are longer on average than reads lacking SD motifs. By failing to isolate the entire

population of ribosome-protected mRNA fragments, Li and Weissman had artificially enriched

for SD containing reads, observing pausing that is not necessarily there in vivo.

Ribosome profiling data can be used to quantify the absolute levels of protein synthesis

genome-wide. In bacteria, many protein complexes are made of components synthesized from a

single operon. In the operon encoding ATP synthase, for example, eight proteins are encoded on

a single polycistronic RNA. The components are incorporated into the complex with various

stoichiometries: 1:10:2:1:3:1:3:143. Li and Weissman showed that bacteria optimize expression

of such complexes by proportional expression, matching the level of expression to the amount

required for complex formation43. This avoids wasteful synthesis of extra protein that would

need to be degraded and raises very interesting questions about how these differences in

translational efficiency are generated.

17

1.3 The Endoribonuclease RelE

In 1983, a mechanism that ensures stable maintenance of plasmids in E. coli was

reported44. When a strain of E. coli cell carrying the F plasmid loses it while dividing, the cell

without the plasmid is not viable44. Only the cells carrying the F plasmid are viable and

continued to divide44. This is because a segment of the F plasmid plays an important role in

maintaining the plasmid in the cell44. This segment, designated ccd (coupled cell division)44, is

composed of two parts: one (ccdB) that expresses an inhibitor of DNA gyrase45 and another

(ccdA) that expresses a protein that inactivates CcdB through forming a protein-protein

interaction44, 45b, 46. When cells inherit the F plasmid, both CcdA and CcdB are expressed and

form complexes that do no harm to the cell44, 46. However, if cells loose the plasmid after

division, the unstable CcdA46 gets degraded rapidly, leaving excess CcdB. Once freed, CcdB

becomes active and arrests cell growth, leading to death upon further division45b, 47. With this

mechanism, the plasmid ensures that it is replicated and passed on to all the viable daughter

cells45b, 47. In this context, the mechanism is called post-segregational killing (PSK)45b, 48 and the

protein module is sometimes called as addiction module49. Later, many such protein pairs were

found on the E. coli chromosome and named as toxin-antitoxin (TA) pairs, since one has

cytotoxic activity and the other antagonizes its pair48.

RelE is a toxin widely found in bacteria and archaea48. As an endoribonuclease, RelE is

known to selectively cleave translating mRNAs by binding to the ribosomal A site and inducing

global translational inhibition50. RelE activity is induced under nutritional depletion and is

closely related to prokaryotic cell survival and pathogenicity including general stress responses

and persister formation50b. RelE is one of the most studied toxins, but its activity and mechanism

are still not well understood.

18

Pedersen et al. showed codon-specific mRNA cleavage by RelE in their in vitro kinetic

assays50a. They observed that UAG and UAA among stop codons and CAG and UCG among

sense codons were the most rapidly cleaved by RelE at the ribosomal A site, normally between

the second and third bases50a. Neubauer et al. solved the structure of 70S ribosomes bound to

RelE; their data partly explain why RelE cleaves mRNA in a codon specific manner and

suggested a general acid-base catalytic mechanism51. They proposed that purines are favored for

the second and third base of A-site codon for better stacking with residues from RelE and 16S

ribosomal RNA in facilitating RelE activity51.

Hurley et al., however, observed different RelE cleavage patterns in vivo52. They found

that RelE cut mRNAs at several sites frequently and efficiently within the first 100 codons of

coding regions, and rarely cut mRNAs at sites near the 3’ end52. They did not see any preference

for CAG or UCG sense codons or any statistically significant sequence preferences. Moreover,

they noticed that some cleavages occurred after the third base of A-site codon while most

cleavages occurred between the second and third as previously known. They argued that their

observance is more consistent with the fact that RelE causes rapid and comprehensive mRNA

degradation and concomitant growth arrest52.

Recently, we found in our work that RelE has some advantages for use as a nuclease in

the ribosome profiling protocol. When added to cell lysates, purified RelE can cleave translating

mRNAs at ribosomal A site and provides better information on ribosome position and reading

frame than does MNase digestion. However, the sequence preference of RelE activity is

problematic and requires further refinement and computational adjustment in analyzing data.

This work will be described in Chapter 2.

19

1.4 tRNA Methylations and TrmD

TrmD is a bacterial-specific S-adenosyl-methionine (AdoMet)-dependent tRNA methyl

transferase. Methylation of tRNA can affect the accuracy of protein synthesis and the

maintenance of reading frame. The N1-methylguanine at position 37 (m1G37)-tRNA product of

TrmD improves the accuracy of protein synthesis on the ribosome. m1G37-tRNA reduces +1

frameshift (+1FS) errors at slippery mRNA sequences and decoding errors of uridine-5-

oxyacetic acid at position 34 (cmo5U34), a wobble modification that frequently accompanies

m1G37 in natural tRNAs53. Loss of TrmD leads to accumulation of +1FS errors leading to pre-

mature termination of protein synthesis at out of frame stop codons. trmD is an essential gene in

E. coli and many bacteria and its biochemical mechanism is very different from eukaryotic tRNA

methyltransferases, leading to interest in TrmD as a drug target50b, 54. We sought to achieve a

cellular-level understanding of how inactivation of TrmD reduces synthesis of proteins to

determine the reasons for its essentiality. In Chapter 3, we report ribosome profiling studies to

observe defects in translation and identify genes whose expression is altered in TrmD deficient

cells.

20

2 MRNA CLEAVAGE BY RELE

2.1 Introduction

Bacteria face enormous selective pressure from their physical and chemical environment

and from competing micro-organisms. In response to this pressure, bacteria have evolved

mechanisms to rapidly regulate gene expression in response to reduction in the levels of

available nutrients, for example, or in response to antibiotics released by other organisms.

Another strategy bacteria use to deal with these stresses is to maintain a small fraction of the

population in a dormant state that survives environmental insults and then resumes growth when

conditions improve55. This strategy plays an important role in antibiotic resistance because

dormant cells (known as persisters) are not killed, even at high antibiotic concentrations.

Shutting down protein synthesis is an essential step for both of these strategies55; for example,

regulating gene expression in response to nutrient starvation via the classical stringent response

and maintaining persister cells in a dormant state as a bet-hedging strategy56.

The relE gene plays a role in stress-response pathways by blocking translation50b. The

gene was originally discovered in genetic screens involving nutrient starvation and was named

for its effect in the stringent response. During amino acid starvation, the concentration of

uncharged tRNAs increases; as these tRNAs bind to the ribosome, they activate the synthesis of

guanosine tetraphosphate (ppGpp) by the RelA protein57. Through its interactions with RNA

polymerase, ppGpp leads to stringent inhibition of rRNA synthesis57. E. coli mutants in which

21

RelE is more highly expressed quickly shut down translation upon amino acid depletion,

stopping the consumption of amino acids and rapidly lowering ppGpp to its pre-starvation levels,

allowing rRNA synthesis to resume after a brief delay58. The synthesis of rRNA during amino

acid starvation is characteristic of the ‘relaxed’ phenotype observed in mutants of genes in

stringent response pathway, hence the name RelE59.

In addition to its role in gene expression, RelE is also a member of the type II toxin-

antitoxin family implicated in persister-cell formation59a. Overexpression of the RelE toxin

causes a reversible inhibition of cell growth resembling the dormant state characteristic of

persister cells59a. Growth resumes when RelE is neutralized by overexpression of its binding

partner, the RelB anti-toxin. Under rich conditions, RelB is expressed at a slightly higher level,

masking RelE activity, but under stress conditions, Lon protease degrades enough of the more

labile RelB anti-toxin to induce RelE activity60. Even under rich conditions, stochastic activation

of RelE and similar toxins is thought to be responsible for inducing a persister-like state in a

small fraction of cells in culture61.

Like at least ten members of the type II toxin-antitoxin family in E. coli, RelE exerts its

effects by cleaving mRNA and inducing translational arrest62. RelE binds in the ribosomal A site

and cleaves the RNA after the second nucleotide in the A-site codon50a. It is an unusual

endonuclease in the sense that it does not cleave RNA outside of this context50a; the ribosome

provides an environment essential to its catalytic activity. A catalytic mechanism has been

proposed based on the X-ray crystal structure of RelE in a 70S ribosome complex and follow-up

kinetic studies51, 63. Initially, RelE was reported to be highly specific for a select codons (CAG,

UCG, and the stop codon UAG) based on its in vitro kinetics50a, but more recent studies suggest

that it has quite a broad specificity in vivo with only a modest sequence preference52. Analyses of

22

cleavage sites on a handful upon highly-expressed genes identified many cleavage sites with

only a modest preference for cleavage before G50a, 52. Woychik and co-workers further made the

puzzling observation that RelE cleaves primarily at the 5’-end of mRNAs, within about the first

100 codons52; the mechanism underlying this polarity is unknown.

Here we report a genome-wide characterization of protein synthesis in E. coli upon RelE

overexpression. Unlike the elegant method of Woychik and co-workers, who adapted RNAseq to

detect cleavage by the MazF toxin64, the ribosome profiling method we employed reports on the

position of ribosomes, allowing us to observe RelE’s effect on translation. We find that ribosome

density is strongly enriched at the 5’-end of genes and propose a model involving cycles of

mRNA cleavage, rescue of stalled ribosomes, and initiation that explains the earlier observation

of preferential RelE cleavage in the first 100 codons52.

Furthermore, we find that RelE can be used to improve the resolution and power of

ribosome profiling in bacteria. As originally developed in yeast, the protocol calls for digestion

of naked mRNA with RNase I to generate ribosome footprints35. As RNase I is inhibited by E.

coli ribosomes, bacterial ribosome profiling studies have used MNase instead. Unfortunately,

MNase is sequence specific, creating strong sequence bias at both the 5’- and 3’-ends of the

RNA fragments, enriching or suppressing specific RNA fragments depending on their nucleotide

sequence. Additionally, MNase generates a broad distribution of the lengths of the ribosome

footprints, unlike the tight 28 nt footprint observed in ribosome profiling studies in yeast42. This

means that the position of the ribosome cannot be determined with sufficient precision to

determine its reading frame42. Addition of purified RelE to cell lysates generates ribosome

footprints that, for the first time, provide excellent information on the position and reading frame

of the ribosome.

23

2.2 Results and Discussion

2.2.1 Ribosomes are Enriched at the 5’-end of Genes in Cells Overexpressing RelE

To study the effect of RelE on protein synthesis in vivo, we overexpressed RelE in wild-

type E. coli MG1655 cells and analyzed ribosome occupancy genome-wide using ribosome

profiling. RelE expression from an araBAD promoter was induced for one hour beginning in

early log phase. Given the high level of RelE expression (one-quarter of ribosome footprints map

to the RelE mRNA) and the long induction period, we observed very pronounced effects on

protein synthesis, presumably representing an equilibrium state in which RelE activity has

exerted its effect and cell growth has slowed or halted. Although strong overexpression no doubt

exaggerates the effects that we observe in comparison to activation of endogenous RelE by

physiological stimuli, the resulting data provide a clear picture of the scope and specificity of this

endonuclease and its effects on gene expression.

The most striking effect of RelE overexpression is the dramatic enrichment of ribosome

occupancy at the 5’-end of genes. In the model of the manX gene shown in Figure 1, for

example, far more ribosome footprints map to the 5’-end of the gene than the 3’-end (red). In

contrast, in the wild-type control (WT1, black), ribosome occupancy is distributed more or less

similarly across the manX gene. These observations hold true genome wide. In a plot of average

ribosome occupancy of about 1000 genes over 1000 nt long that were aligned at the start codon,

ribosome coverage remains fairly constant in the wild-type control, whereas in the RelE1 data

ribosomes are highly enriched in the first 400 nt and depleted after the first 400 nt. (Note that the

vertical spread in the RelE1 data arises from the reading frame, as will be discussed below). In

addition, a very strong peak of ribosome density is observed at the start codon in the RelE1 data.

These data show that as a result of RelE activity, ribosomes accumulate at the 5’-end of genes

24

and are depleted at the 3’-end. This suggests that the number of ribosomes completing the

synthesis of full-length proteins is strongly reduced, consistent with RelE’s known function of

inhibiting protein synthesis and arresting cell growth.

Given that RelE is an endonuclease that cleaves mRNA inside the ribosomal active

site50a, we also looked at steady-state levels of mRNA in RNAseq libraries prepared from the

WT1 and RelE1 samples. Here again, we see enriched RNA density at the 5’-end of the manX

Figure 1 Average ribosome occupancy

Ribo-seq (left) and RNAseq (right) reads at the manXYZ operon (top). Average ribosome occupancy at

about 1000 genes aligned at their start codons (bottom).

25

gene when RelE is overexpressed (Figure 1, top right, blue), whereas the coverage is relatively

uniform in the wild-type control (grey). When the RNAseq from many genes is averaged

together, a modest two-fold increase in density at the 5’-end is observed in the RelE1 data

(Figure 1, bottom, blue). These findings are consistent with the ribosome profiling data in

showing a 5’ to 3’ polarity.

Finally, a careful inspection of gene expression across polycistronic transcripts suggests

that ribosome density decays not only across single genes but across operons as well. The

manXYZ operon, for example, is expressed as a single transcription unit from a single promoter,

a fact supported by equal levels of RNA density for the three genes in the WT1 RNAseq dataset

(Figure 1, top right). Translation of all three genes is also observed in the WT1 Ribo-seq data

(top left, black). In contrast, lower levels of RNA and far lower levels of ribosome footprints are

observed for the downstream manY and manZ genes compared with manX.

To quantify this effect, we compared the levels of about 2500 genes in the RelE1 and

WT1 samples; these ratios are shown separated by the position of the gene within its operon

(Figure 2). The transcriptome of E. coli is remarkably complex: many operons overlap and often

multiple promoters regulate subsets of genes within an operon. To define single transcription

units, we started with roughly 4,000 annotated operons in the RegulonDB database and excluded

those in which the genes varied in RNAseq density by more than five-fold (using RNAseq data

from the WT1 sample). Like the manX gene, which has three-fold more ribosome footprints in

the RelE1 sample than in the WT1 sample, most genes that are at the first position in an operon

or are monocistronic have more ribosome footprints in the RelE1 sample. In contrast, genes that

lie downstream on the operon have fewer ribosome footprints. The manZ gene has two-fold

26

fewer reads in the RelE1 sample

than in the WT1 sample (Figure

1). The distribution for genes at

the first position versus genes in

the fifth position or further

downstream is statistically

significant (Mann-Whitney P-

value of 8.4 x 10-22). These

differences are also observed at

the RNA level, although

perhaps to a lesser extent. In

summary, our data indicate that

upon RelE overexpression, ribosomes are enriched not only at the 5’-end of the units of

translation (open reading frames) but at the 5’-end of units of transcription (polycistronic

mRNAs).

We propose the following model for enrichment of ribosomes at the 5’-end of genes upon

RelE expression. As RelE begins to accumulate in the cell, it enters the A site of ribosomes and

cleaves mRNA, preventing further cycles of elongation. Stalled ribosomes are rescued by

tmRNA-SmpB or by the backup system comprised of the ArfA protein and RF2. Several

biochemical studies have shown that tmRNA reacts particularly well with ribosomes following

the loss of mRNA downstream of the A site upon RelE cleavage65. Both the tmRNA and ArfA

systems release the nascent peptide from the stalled ribosome and promote recycling of the

subunits. The free subunits are then free to assemble on another transcript at the start codon and

Figure 2 Ribosome enrichment at 5’-end genes of operon

Ribosome occupancy (red) and RNAseq density (blue) are

enriched in genes at the 5’-end of operons upon RelE

overexpression.

27

begin elongation again. If RelE cleaves the initiation complex, a strong start codon peak is

observed. If a few cycles of elongation take place before cleavage of the mRNA, ribosomes

move into the coding sequence away from the start codon. In this manner, cycles of mRNA

cleavage, ribosome rescue, and initiation could enrich for ribosomes at the 5’-end of genes.

It is possible that RNA decay mechanisms contribute to this effect as well. Following

cleavage by RelE, the 3’-end of the upstream RNA fragment is protected from degradation by

exonucleases by the stalled ribosome. The higher ribosome density on the upstream fragment

could also prevent degradation by endonucleases such as RNase E. In addition, the upstream

fragment has a 5’-triphosphate which is stabilizing. In contrast, the downstream fragment may be

less able to recruit ribosomes and will have a 5’-monophosphate that helps to recruit RNase E

promoting further endonucleolytic cleavage and RNA decay.

In support of this model, we note that in a previous study of RelE cleavage in highly

expressed genes using primer extension, cleavage sites were over-represented at the 5’-end of

genes, although no explanation for this observation was provided. This observation is related but

subtly different from ours: we observed enrichment of ribosomes at the 5’-end of genes, and they

observed high levels of RelE cleavage at the 5’-end of genes. The reloading of ribosomes at start

codons followed by a few rounds of elongation prior to additional cleavage events could tie these

observations together: RelE is targeted to the 5’-end of genes because that is where the

ribosomes accumulate over time.

As further evidence of the importance of rescuing stalled ribosomes, we note that some of

the rescue machinery is strongly upregulated in the RelE1 sample. Steady-state levels of the

backup rescue factor ArfA are 15-fold higher at the RNA level in the RelE1 sample. The levels

of synthesis of the ArfA protein are more than 80-fold increased. ArfA expression is regulated by

28

tmRNA in an elegant feedback mechanism: the transcript is cleaved by RNase III within the

open reading frame, leading to ribosome stalling and production of a truncated protein product.

tmRNA rescues these ribosomes stalled during ArfA synthesis and tags the truncated product for

proteolysis. If the capacity of tmRNA is exceeded, ArfA begins to accumulate through a poorly

understood drop-off mechanism. The fact that ArfA is strongly upregulated in cells treated with

RelE indicates that the tmRNA system is overwhelmed and that cells are responding for the need

for more ribosome rescue activity by strongly upregulating ArfA expression. Taken together, our

findings support a model in which mRNA cleavage by RelE, ribosome recycling, and initiation

enrich ribosome occupancy at the 5’-end of genes.

2.2.2 RelE Predominantly Cleaves mRNA after the Second Nucleotide in Empty A sites

The ribosome profiling data obtained from cells overexpressing the RelE endonuclease

provide information about cleavage sites genome-wide. In a sense, this information is indirect;

we measure ribosome footprints and not the products of RelE cleavage themselves. In preparing

Figure 3 Average ribosome density at start- and stop-codon

Average ribosome density at start (left) and stop codons (right) for Ribo-seq data from wild-type

cells or cells overexpressing RelE.

29

the footprints, another nuclease (MNase) is used to digest mRNA unprotected by ribosomes.

Although the MNase treatment could potentially interfere with our ability to detect RelE

cleavage events, we find that for the majority (if not all) of the footprints, the 3’-end is generated

by RelE cleavage and not MNase. Perhaps this is due to the high concentration of RelE in the

cell over the hour-long induction period.

The fact that most ribosome footprints are generated by RelE is clearly evident in the

position of the peak at the start codon in plots of average ribosome occupancy (Figure 3, left).

During initiation, the AUG start codon is positioned in the ribosomal P site. In the WT1 control

sample, MNase digests mRNA back to the 3’-boundary of the ribosome, 15 nt downstream of the

first nucleotide of the start codon (black). Although the start codon peak at +15 is small in the

WT1 sample, its position matches what we previously observed in many bacterial profiling

libraries generated with MNase. The position of the ribosome is most accurately and precisely

determined from the 3’-end of footprints in bacterial ribosome profiling; plots of ribosome

density reflect the 3’-end of sequencing reads. Furthermore, this position is consistent with the

distances observed in toeprinting assays in which reverse transcriptase is blocked by the 3’-

boundary of the ribosome and arrests 15-16 nt downstream of the first nucleotide in the P site

codon66. In contrast, the start codon peak in the RelE1 data is four nt downstream of the first

nucleotide of the start codon (red). In other words, the mRNA is cleaved between the second and

third nucleotide in the A-site codon within initiation complexes. The fact that the signal is far

higher at +4 than at +15 means that most if not all of the ribosome footprints are cleaved by RelE

at their 3’-ends.

The same phenomenon is observed at peaks at stop codons in plots of average ribosome

occupancy (Figure 3, right). During termination, the stop codons UAG, UGA, or UAA are

30

positioned in the ribosomal A site. In the WT1 sample, MNase digests back to the 3’-boundary

of the ribosome, 12 nt downstream of the first nucleotide in the stop codon (black). In contrast,

RelE cleaves after the second nt in the stop codon (red) in the A site. Together, the data at start

and stop peaks show that RelE cleaves predominantly after the second nucleotide in the A-site

codon and that most ribosome footprints were generated by RelE cleavage at their 3’-ends. This

pattern of cleavage is consistent with previous enzymatic, structural, and primer extension

studies of RelE activity.

In examining the peaks at other well-characterized ribosome pausing sites, we observed

that the ribosomal A site must be empty for RelE cleavage to occur. The SecM and TnaC

polypeptides interact with the ribosome to inhibit their own synthesis in response to specific

cellular signals67. The ribosome stalls at the RAGP sequence in SecM, for example, with the Gly

codon in the P site and unreactive Pro-tRNA bound in the A site67a. A strong peak 15 nt

downstream of the Gly codon in the WT1 sample corresponds to this well-understood pausing

event when the samples are treated only with MNase (Figure 4, left). In the RelE1 data, the

expected peak after the second nucleotide in the Pro codon is not observed; instead, the strongest

Figure 4 Ribo-seq data at nascent peptide-mediated stalling sites

SecM (left) and TnaC (right).

31

peak overlaps with the one observed in the WT1 sample 15 nt downstream of the Gly codon.

This suggests that RelE is unable to cleave the Pro codon in the A site in stalled SecM

complexes, presumably due to the Pro-tRNA trapped and blocking RelE entry. Similarly,

ribosome stalling at the C-terminus of the TnaC peptide (Figure 4, right) is observed in the WT1

data with a strong peak 15 nt downstream of the final residue, Pro67b. In the RelE1 data, a peak is

expected in the stop codon positioned in the A site, but none is observed, only the same peak

observed in the WT1 data. RelE is probably blocked by trapped release factor 2 that is recruited

but cannot complete hydrolysis of the nascent peptide. These findings suggest that RelE is

capable of competing with aminoacyl-tRNA and release factors for access to the A site during

normal elongation and termination, but that factors that are trapped long-term during

programmed ribosomal arrest block RelE activity.

2.2.3 in vitro Digestion with RelE Reveals the Reading Frame

It has been impossible to detect reading frame in ribosome profiling studies in bacteria

and the MNase enzyme is at least partly to blame, leaving one or more nucleotides undigested at

the 3’-boundary of the ribosome. Given the precision with which RelE cleaves after the second

nucleotide in the A site when expressed in vivo, we hypothesized that RelE might prove to be a

useful tool in ribosome profiling, generating more reliable 3’-ends and improving the resolution

at which the position of the ribosome can be determined. We purified RelE and added it to cell

lysates under our regular digestion conditions in vitro; MNase was included at the regular

concentrations to digest mRNA to the 5’-boundary of the ribosome. Ribosome footprints 10 – 40

nt in length were then cloned and sequenced following our usual protocol.

32

Comparison of profiling libraries obtained by expression of RelE in vivo (RelE1, Figure

1) with those obtained by in vitro RelE digestion (RelE2, Figure 5) reveals many similarities but

also an important difference. The activity of RelE in vitro can be seen in plots of average

ribosome occupancy across many genes aligned at start codons. A strong start codon peak in the

RelE-treated sample (RelE2) is found at the +4 and not the +15 site (Figure 5). Furthermore, like

the RelE1 data in Figure 1, the vertical spread in the RelE2 signal is substantially larger than that

observed in a library prepared from the same biological sample using only MNase (WT2),

indicative of 3’-periodicity reflecting reading frame. On the other hand, we do not observe the

marked decay of ribosome density at the 3’-end of genes that we observe when RelE is

overexpressed in vivo (see the data from the RelE1 library in Figure 1). This is because

translation is blocked in the cell lysate; elongation is arrested by chloramphenicol and initiation

is probably inhibited by dilution of the necessary factors and exhaustion of GTP. This is further

Figure 5 Average ribosome occupancy for a wild-type sample

RNA was digested in vitro with MNase alone (black) or with purified RelE protein (red).

33

evidence that the accumulation of ribosome occupancy at the 5’-end of genes in vivo is due to the

dynamics of cleavage, rescue, and initiation and not RelE activity per se.

Returning to the question of reading frame, we calculated the average ribosome density at

all three sub-codon positions in open reading frames throughout the genome. The library

prepared with MNase (WT2) shows slightly higher density at the first nucleotide (40%) than the

other two nucleotides (30% each, Figure 6, top left). Given that this effect is lacking in footprints

mapping to 3’-

untranslated regions (3’-

UTRs) (where the reading

frame is defined as the

same as the preceding

ORF), it may be tempting

to attribute this 3 nt

periodicity to the

ribosome’s reading frame

(bottom left). We found,

however, that RNAseq

libraries prepared by

digestion of total RNA by

low concentrations of

MNase yielded the same

result—ribosome occupancy is enriched at the first sub-codon position in ORFs but not in 3’-

UTRs. The sequence specificity of MNase coupled with the nucleotide bias in ORFs explain this

Figure 6 Ribo-seq and RNAseq density at each position

Ribosome density within codons in coding sequences (top) or 3’-

untranslated regions (bottom). Right: sequence bias in the genome

in the CDS or 3’-UTR.

34

observation. MNase cleaves preferentially before A and T which are enriched at position two in

codons in ORFs but not 3’-UTRs (Figure 6, right). Cleavage before position two makes the 3’-

end of footprints align with the first sub-codon position. Given that ribosome occupancy is

assigned using the 3’-end of footprints, ribosome density appears higher at the first position

(40%) than the other two (30% each).

In contrast, the libraries prepared by RelE digestion in vitro show the strongest density at

sub-codon position two (Figure 6, top left), corresponding to cleavage before the third nucleotide

in the A site codon, as observed at both start and stop codons in average ribosome density plots

(Figures 3 and 4). Roughly 60% of footprints map to the second sub-codon position; 30% map to

position three while only about 10% map to position one. The large difference between the

density at the first two positions explains the large vertical spread in the RelE2 data observed in

Figure 5 and provides the first reliable information about ribosome reading frame in bacterial

ribosome profiling data. Footprints from the 3’-UTR show no evidence of reading frame,

consistent with the expectation that ribosomes found there are not synthesizing protein or

moving in 3 nt steps (Figure 6, bottom left).

2.2.4 RelE Prefers to Cleave after C and before G

By analyzing the 3’-ends of ribosome footprints, we are able to observe the specificity of

the MNase and RelE enzymes. As others have shown, MNase preferentially cleaves before the

nucleotides A and T; in our data we observe that A and T are strongly enriched downstream of

the cleavage site, both +1 nt and to a lesser extent +2 nt as well (Figure 7, left). In contrast, A

and T are underrepresented upstream of the cleavage site, both −1 nt and to a lesser extent −2 nt.

The sequence preferences of RelE are very different from those of MNase: C is preferred at the

−1 position while G is selected against (Figure 7, right). Following the cleavage site, G is

35

strongly preferred and C is selected against. There appears to be a slight avoidance of A and T at

both positions. These tendencies hold true whether RelE digestion occurs in vitro or in vivo. In

contrast with what we observed with MNase, the sequence two or more nucleotides away from

cleavage site exert a small effect.

Although initial reports of RelE activity described the endonuclease as highly specific for

a few codons, additional studies revealed a more relaxed specificity. The preference for G after

the cleavage site was anticipated by the kinetic studies of Ehrenberg and co-workers, who found

that kcat/KM values for RelE cleavage were markedly higher for codons ending in G50a. The

presence of G at the third position of the codon had the strongest effect in explaining their kinetic

data. Woychik and co-workers also reported a strong preference for cleavage before G in five

highly expressed genes in E. coli52. Finally, the X-ray crystal structure of RelE bound in the A

Figure 7 Sequence bias at the 3’-end of ribosome footprints

Different cleavage patterns due to cleavage activity by RelE both in vivo and in vitro. All samples

were also digested with MNase in vitro.

36

site of 70S ribosomes offers a possible explanation of this preference: the G in the third position

of the A site codon stacks on the base of C1054, a conserved 16S rRNA nucleotide in the

decoding center. The structure suggests that direct contact of RelE residues with the G nucleotide

are also possible.

In contrast to the good agreement of our data with these earlier studies regarding the

specificity of the downstream nucleotide at the cleavage site, our observations of a marked

preference upstream of the cleavage site are unexpected. There is little evidence in the literature

of a strong bias for C and against G at the −1 position, raising the possibility that this effect is an

artifact of ribosome

profiling, arising from

biases in cloning

ribosome footprints.

We find, however,

that RNAseq libraries

show no such bias

when fragments were

prepared by alkaline

hydrolysis and then cloned and sequenced exactly as the profiling libraries. This finding rules out

the possibility that the preference for C is the result of cloning bias and strongly implicates RelE

cleavage itself as the cause, given that the method of RNA digestion is the only difference

between the two protocols.

Figure 8 Sequence bias at the 3’-end of RNA fragments

in two libraries. The fragmentation was done by alkaline hydrolysis.

37

2.2.5 The Specificity of RelE Interferes with Analyses of Ribosome Pausing

Although the generation of ribosome

footprints with RelE allows us to determine

the ribosome position and reading frame

with high resolution, the use of RelE incurs

certain disadvantages as well. Both MNase

and RelE are sequence-specific nucleases, a

property that creates bias at the ends of

ribosome footprints. As shown in Figure 9

(top), plots of average ribosome occupancy

at sense codons are noisy when the 3’-end of

the reads line up with the codon of interest

(that is, the distance from the codon is close

to zero). This noise arises from enrichment

of certain sequences at the 3’-end of reads at

the specific nucleotides in the codon being

averaged. For example, there is a strong

peak at 0 in the GTA codon (green) because

cleavage after G and before T is optimal

given the specificity of MNase. In contrast,

there is a strong peak at 1 for the CCA

codon (blue) because cleavage after position

2 is preferred. Although the sequence bias at

Figure 9 The ribosome occupancy on codons

Top: Average ribosome occupancy at all 61

sense codons. The position corresponding to

the codon in the A, P, or E site is indicated.

Middle: The same plot with data from the RelE

in vitro digest. Bottom: the differences in pause

score at all 61 sense codons between ribosome

profiling libraries.

38

the 3’-end of reads skews the data, in practice this noise can be ignored because the noise that

this creates is at the 3’-boundary of the ribosome, far away from functional sites such as the A, P,

and E sites for tRNA-binding. When the 3’-end of reads are 15 nt downstream of codons of

interest, positioning the CCA and GTA codons in the P site, for example, noise from cloning bias

does not interfere.

In contrast, the pronounced sequence specificity of RelE creates noise that obscures the

signal right at functional sites in the ribosome. Because RelE cleaves mRNA in the ribosomal A

site, the noise generated by bias at the 3’-end exactly overlaps the A site, making it difficult to

draw conclusions about the dwell time of ribosomes on specific codons during elongation or

termination (Figure 9, middle). We computed pause scores for three ribosome profiling libraries;

two biological replicates in which the footprints were generated by MNase (WT2 and WT3) and

the third library from the same biological sample as the WT2 library but using both RelE and

MNase (RelE2). We observed relatively small differences in the pause scores between the WT2

and WT3 libraries, in spite of the fact that they were derived from independent biological

samples (Figure 9, bottom). This holds true for all three tRNA-binding sites. There were

significant differences, however, between the levels of pausing in the WT2 and RelE2 data, even

though they were derived from the same biological sample. The differences were more

pronounced in the A site than the P or E sites. We conclude that in vitro digestion by RelE

introduces noise that complicates the analysis of ribosome pausing and argue that the MNase

data are more likely to represent an accurate view of pausing in vivo given that the cloning bias

in the MNase libraries is farther away from the sites of interest.

39

2.2.6 Refined RelE-Derived Ribosome Density Better Reflects Reading Frame

The sequence specificity of RelE alters the pattern of cleavage in predictable ways; by

taking this into account, we can obtain even better information regarding reading frame. Looking

in more detail at average ribosome occupancy at sense codons positioned in the A site, we find

that most codons are cleaved after the second nucleotide, as expected from the analyses of start

and stop codons above. In Figure 10 (left), the ribosome occupancy has been shifted such that the

A-site codon starts at zero; the peak at 1 observed in most codons (grey) corresponds to cleavage

after the second nucleotide. A subset of codons has higher density at position 2 corresponding to

cleavage after the third nucleotide in the codon (Figure 10, red): all of these codons end in C.

Given the specificity depicted in Figure 7, where cleavage after C is preferred and cleavage

before C is strongly inhibited, it makes sense that NNC codons are cleaved after the third

position, between the A site codon and the codon downstream. To compensate for this cleavage

bias, we shifted the ribosome density at all NNC codons from the third nucleotide to the second.

This improved the signal at the second position from about 60% to about 80%, with 10%

remaining at position one and 10% at position three (Figure 10, right). This suggests that two-

Figure 10 Correction for better reading frame

Left: average ribosome occupancy on all sense codons, shifted from Figure 9 so that the A site

codon starts at zero. Right: reading frame when the reads on NNC are shifted from position 3 to

position 2.

40

thirds of reads that align to the third sub-codon position arise from NNC codons. Given what we

know about sequence specificity a more sophisticated algorithm could be employed in the future

to realign ribosome density to optimize reading frame.

Even by simply shifting density on NNC codons from the third to second position,

reading frame is significantly improved. With this level of resolution, frameshifting events can

be detected in bacterial ribosome profiling data, as can be seen in analysis of the programmed

frameshift on the prfB gene encoding RF268. When RF2 levels are limiting, ribosomes pause at a

stop codon at the 28th codon in the gene and then shift into the +1 frame. Following the

frameshift, ribosomes complete the synthesis of the RF2 protein which is encoded in the new

reading frame. When we split the ribosome occupancy signal into three components, with reads

that map to the first, second, or third sub-codon position, this frameshifting behavior is evident in

models of the prfB gene. Upstream of the frameshift site, the majority of ribosome footprints are

cleaved after the second sub-codon position, as expected. But downstream of the frameshifting

event, most footprints are cleaved after the third position, consistent with a +1 frameshift. Very

few reads map to the first sub-codon position, either before or after the frameshift site.

Figure 11 Programmed frameshifting at the prfB gene

Left: ribosome occupancy split into three component parts, by sub-codon position. Given RelE’s

preference for cleaving after the second nt, the blue data represent in-frame translation upstream of

the frameshift site and the red data represent in-frame translation downstream (in the +1 frame).

41

Quantitation of the reading frame upstream and downstream of the programmed frameshift site

further supports this conclusion. Going forward, the high degree of three nucleotide periodicity

arising from translational reading frame can be used to search for other programmed

frameshifting sites in bacteria, as has been done in human ribosome profiling studies37 .

2.2.7 Concluding Thoughts

Our ribosome profiling analyses of RelE activity in vivo reveal dynamic cycles of

cleavage, ribosome rescue, and initiation that enrich ribosomes at the 5’-end of genes. Although

it was previously known that RelE-cleaved mRNAs are good targets for the tmRNA rescue

system, our data highlight the importance of the tmRNA and ArfA systems in the cellular

response to and recovery from RelE activity. How cells exit the dormant state with high RelE

activity is not yet clear, but it seems likely that ribosome rescue is essential for this to occur.

Antibiotics are beginning to be developed to target the tmRNA rescue pathway, raising the

possibility of targeting persister cells due to their reliance on toxins to cleave mRNA and block

translation.

The use of RelE as a nuclease for ribosome profiling suffers from the same problem of

sequence specificity as the enzyme normally used, MNase. This makes the enzyme less than

ideal for generating libraries where careful measurement of ribosome pausing is the goal.

However, RelE cleaves the 3’-end of fragments more precisely, allowing us to observe the

ribosome’s reading frame for the first time in bacteria.

In the future, it may be possible to generate ribosome profiling libraries using RelE alone.

In our in vitro digests, we also added MNase to cleave the mRNA back to the 5’-end of the

ribosome generating a ribosome footprint. With RelE alone, cleavage would only occur in the A

site, generating fragments that are far longer than the 10 – 40 mer ribosome footprints that we

42

work with now. In theory, these fragments could be cloned without size selection. The distance

between ribosomes on messages could then be determined by sequencing both ends using 50 bp

paired Illumina sequencing. This might give insights into ribosome stacking at strong pausing

sequences and a picture of ribosome distribution on single messages, since the sequence between

two ribosomes would remain intact.

2.3 Materials and Methods

2.3.1 Bacterial Strains and Growth Conditions

E. coli K-12 strain MG1655 was used as a wild-type strain. The RelE overexpression

strain was constructed by transformation with a plasmid, pJC203, which contains an araBAD

promoter. Cell cultures were grown at 37°C in either LB media or MOPS media supplemented

with 1% glucose, all 20 amino acids, and other nutrients (Teknova) and RelE overexpression was

induced by adding arabinose (final 0.2%) at an O.D600 of 0.2-0.3. Cell cultures were grown

further either for 1h or 20 min before harvesting.

2.3.2 Preparation of Ribo-seq and RNAseq Libraries

Ribo-seq libraries were generated as described previously66 with some modifications.

Ribosome-protected fragments were size-selected between 20 – 40 nt in length initially, but 10 –

40 nt in the latter libraries in an attempt to include short fragments generated by RelE cleavage

activity. For the library digested by RelE in vitro, 1 nmol of purified RelE was also added when

25 AU of RNA in the lysate was digested with 3000 U of MNase (Roche) for 1 h at 25 °C.

43

RNAseq libraries were also size-selected between 10 – 40 nt in length in the latter

preparation to reveal 3’-end sequences, which was not available due to longer reads size

selection with the length between 40 – 60 nt.

2.3.3 Data Analysis

Data analysis obtained from Illumina sequencing were described previously66.

44

3 THE ROLE OF TRNA METHYLTRANSFERASE TRMD

This work was a collaboration with Isao Masuda of Ya-Ming Hou’s lab at Thomas

Jefferson University. Dr. Masuda created and characterized the TrmD depletion system and his

work is described here for completeness. My contribution was to assist him in optimizing the

growth conditions, generate the ribosome profiling libraries, and analyze and interpret the

profiling sequencing data.

3.1 Introduction

tRNAs are adaptor molecules that deliver amino acids to the ribosome for protein

synthesis. To ensure the fidelity of translation, each tRNA must carry its cognate amino acid and

decode mRNA using the correct match between its anticodon and the codon in mRNA. tRNAs

undergo numerous chemical modifications after their transcription; an average E. coli tRNA

contains 77 nucleotides and have about 8 of them modified. These modifications play an

important role in the fidelity of both aminoacylation and decoding.

Most tRNA modifications are located near the anticodon69. One of the purposes of

modification is to fine tune the anticodon so that a single tRNA can read multiple, degenerate

codons. There are 47 different tRNAs produced from 86 genes in E. coli; these tRNAs are able to

decode all 61 sense codons. tRNA modifications promote this flexibility in codon recognition

while at the same time maintaining fidelity. This is a difficult but critical balance to achieve: the

45

tRNAs must be allowed a certain amount of broadening of specificity but only under the right

circumstances.

tRNA modifications also play a critical role in maintaining the reading frame during

translation46. One particularly important enzyme for preventing frameshifting is TrmD, a

bacterial enzyme responsible for methylation of N1 of G37 of several tRNAs70. Unlike the case

with other modifications, deficiency of m1G37 leads to cell death in many bacteria including E.

coli70b, 71. TrmD is very different structurally and mechanistically from its eukaryotic homolog,

Trm572. Studying E. coli TrmD can give us a framework for understanding methyltransferases,

the role of modifying enzymes in tRNA stability and activity, and insight into a promising

antibiotic target. Here we report our efforts to understand the role of TrmD in living cells using

ribosome profiling.

3.1.1 tRNA Synthesis and Post-Transcriptional Modifications

In E. coli, tRNAs are transcribed as long precursors73. Various RNases process the

precursor tRNAs (pre-tRNAs) to mature, functional tRNAs by cutting excess residues at the 5’-

and 3’-ends74. For most pre-tRNAs, RNase E removes nucleotides at the 3’-end by an

endonucleolytic cleavage, leaving only a short 3’-trailer75. Then, RNase P interacts with the CCA

motif and removes nucleotides at the 5’-end of the pre-tRNA with a single endonucleolytic

cleavage; this creates the mature 5’-end of the tRNA. Finally, exonucleases such as RNase T,

PH, II, and D further trim the remaining 3’-trailer nucleotides up to the mature CCA motif at the

3’-end of the tRNA73.

All tRNAs have L-shaped tertiary structures formed by base pairing and coaxial stacking

of the helices. Each tRNA is slightly different but shares common secondary structural

features76. When depicted in two-dimensions, tRNA have a cloverleaf structure consisting of an

46

acceptor stem, D (dihydrouridine) loop, anticodon loop, and T loop. The acceptor stem contains

the conserved 3’-terminal CCA motif where the amino acid is covalently linked77. The structure

of the anticodon loop presents the anticodon in a way that allows it to form the correct base pairs

during decoding77. The other structures contribute to interactions with the translation machinery,

including the ribosome, aminoacyl-tRNA synthetases (aaRSs), and EF-Tu.

Following transcription, tRNAs are modified at several sites to enhance their cellular

functions, promoting their structure and stability, optimizing translational fidelity and accuracy,

and maintaining the correct reading frame. The first modification identified was pseudouridine

(Ψ), as the fifth nucleotide, which is an isomer of uridine78. Ψ is present in the T loop that is

recognized by aaRS’s and EF-Tu. Likewise, the dihydrouridine modification is found in the D

loop79. With two extra hydrogens added to its pyrimidine ring, dihydrouridine is no longer planar

or aromatic, disturbing base stacking, and destabilizes the structure, conferring flexibility79.

Most critical modifications of tRNA, however, occur in the anticodon loop. The first two

positions in a given codon base pair (in classical Watson-Crick fashion) with the second and

third positions of the anticodon, nucleotides 35 and 36 in the tRNA. The third nucleotide in the

codon in the ‘wobble’ position, however, can participate in a number of different pairing

geometries in its interaction with position 34 of the tRNA. Modification of position 34 can

broaden base pairing80: deaminating A34 to inosine, for example, in tRNAArg in bacteria allows

for pairing with A, C, and U at the wobble position. Alternatively, modifications of U34 to 5-

methoxycarbonyl-methyl-2-thouridine (mcm5s2U34) and the 5-methylation of cytosine 34

(m5C34) allow a single tRNA isoacceptor to read multiple codons81. The modifications promote

accurate codon-anticodon interactions, optimizing translation of the genetic code.

47

tRNA modifications also play a critical role in maintaining the reading frame of

translation80, 82. In most cases, frameshifting is far more detrimental than other decoding errors.

Missense errors that replace one amino acid for another often have little or no effect on the

stability or activity of the protein product. On the other hand, most shifts in reading frame lead to

pre-mature termination at a stop codon in the new reading frame, yielding truncated and inactive

polypeptides. Pseudouridine and 2-thiouridylation (s2U34) play roles in frame maintenance in

addition to their other functions78a, 83. One of the best characterized modifications involved in

frame maintenance is the methylation of G37, just downstream of the anticodon.

3.2 The TrmD Methyltransferase

The m1G37 modification involves a very ancient

pathway; proteins catalyzing G37 methylation are found in

all three domains of life84. m1G37 is found in most tRNAs

that read codons starting with C. It is thought that

methylation of G37 may block base pairing between this

nucleotide and the mRNA that would promote decoding of a

quadruplet instead of a triplet (Figure 12)84. Codons starting with C are read by tRNAs with G36;

the combination of G36 and unmethylated G37 is thought to be problematic. Other mechanisms

involving structural changes in the tRNA or faulty translocation on the ribosome may also play a

role.

There are many enzymes that facilitate tRNA methylation. Some enzymes add methyl

groups to carbon or nitrogen atoms in the bases, while others methylate oxygens in the ribose

sugars. In general, methyltransferases catalyze the transfer of a methyl group (CH3) from the

donor to a substrate molecule. In most cases, the methyl donor is S-adenosyl-L-methionine

Figure 12 1-methyl-Guanosine

48

(AdoMet)85. AdoMet-dependent methylation often involves SN2 substitution reactions in which

the substrate attacks on the reactive methyl group in AdoMet, yielding a single stereoisomeric

product.

In bacteria, methylation of G37 is accomplished by the TrmD enzyme, which is

composed of 255 amino acids (28.4 kD) and acts as a homodimer. The active site is formed by

residues from each of the monomers86. The structure contains a unique trefoil knot deep within

each of the homodimers whose function is not fully understood. When AdoMet binds to the

structure, it causes conformational changes. The adenosine moiety of AdoMet is surrounded by a

loop and AdoMet is in a bent conformation85.

It has been proposed that the TrmD homodimer requires the anticodon loop of tRNA to

be precisely positioned near AdoMet for the methylation of G37 since bound AdoMet is located

in a deep, seemingly inaccessible pocket near the center of TrmD85. The conserved

phenylalanine residue is proposed to provide critical stacking interactions required for the final

positioning of G3786. It has been proposed that an aspartate of TrmD acts as a general base while

other conserved residues provide G37 specificity for the reaction allowing methylation of the N1

position of G via a deprotonation, followed by the transfer of the methyl group of AdoMet85.

TrmD binds to tRNAs non-specifically searching for its recognition element, the

dinucleotide G36-G3786. Once TrmD recognizes the two consecutive G residues at the 3’ end of

the anticodon, AdoMet-dependent methylation occurs86. tRNA is positioned in the cleft and, if

the identity element G36pG37 is present, G36pG37 might be ‘flipped’ into the dual pocket

catalytic center for subsequent methylation by bound AdoMet85-86. If a G is not found at position

36, then a stable complex of G37 near AdoMet cannot form and catalysis does not proceed86.

49

The bacterial TrmD enzyme is substantially different from its eukaryotic homolog, Trm5,

raising the possibility of targeting TrmD with antibiotics87. The two enzymes show no detectable

similarity in sequence or structure. While TrmD is an obligate dimer, Trm5 is a monomer88.

TrmD binds AdoMet with the rare trefoil knot, Trm5 binds AdoMet in the open space of a

dinucleotide-fold. As noted above, TrmD binds AdoMet in a bent conformation, where the

adenosine and methionine moieties are bent toward each other by 90° whereas Trm5 binds

AdoMet in the more common straight conformation, with an 180° angle between the two

moieties86, 89. The fact that these enzymes have different active site geometries and both bind

small molecules makes them attractive drug targets, and pharmaceutical companies are working

on developing small molecule inhibitors of TrmD84. Loss of TrmD activity is lethal in E. coli;

this is unusual for tRNA modification enzymes, most of which can be deleted with little or no

detectable phenotype. Although increased levels of frameshifting are observed when TrmD and

m1G37 are lost, it is not entirely clear what the mechanism of cell death is. To obtain a better

understanding of protein synthesis in living cells upon depletion of G37 methylation, we

performed ribosome profiling.

3.3 Results

3.3.1 Achieving Regulation of TrmD, an Essential Protein

An inducible TrmD-depletion E. coli strain was constructed to observe defects in

translation and identify genes whose expression is altered when m1G37 of tRNA is depleted in

cells. Since m1G37 is an essential modification of several tRNAs and deletion of trmD is lethal,

we constructed a strain in which TrmD can be conditionally depleted, starting with the G78

strain90. In G78, proteolysis of certain target proteins can be regulated by controlling the

50

expression of the peptide recognition protein, ClpX, which binds to hydrophobic C-terminal

peptide tags, and Clp, a protease. The ClpXP complex degrades proteins tagged by tmRNA as

well as many other proteins with C-termini recognized by the complex91. In the G78 strain, the

ClpXP genes are regulated by an arabinose inducible promoter, allowing us to regulate its

expression by changing media conditions90. We also express TrmD from a constitutive promoter

with the tmRNA tag translationally fused to its C terminus; this effectively targets the protein for

degradation when ClpXP is expressed upon induction with arabinose.

To construct the TrmD-depeletion strain, the trmD gene, which is expressed with another

three genes on an operon, is modified at its 3’-end to contain a peptide tag (Figure 13). This tag

contains a sequence that targets it for degradation by ClpXP (YALAA) as well as His6 and

FLAG epitopes (Figure

13)90. This is called the

degron strain (with the

Deg-tag); the control

strain also contains the

epitope tags but not the

YALAA degradation

signal (the Cont-tag).

The trmD mutations

were first established by

λ red recombination in

E. coli strain SM140590. Then, the trmD alleles with the added degradation tag or control tag are

Figure 13 TrmD depletion strain construction

51

verified by PCR, and then the alleles were transduced using P1 phage in the G78 strain in which

ClpXP can be regulated, generating the trmD-deg and trmD-cont strains (Figure 13).

The two strains were tested by growing on agar plates (Figure 14, top). When cells were

plated on media containing glucose, both strains formed colonies (data not shown). However,

when cells were grown on the plate with 0.2%

arabinose added, there were significant

differences in colony formation. The degron

strain (trmD-deg) formed fewer colonies than

the control strain, suggesting that the tagged

TrmD was successfully depleted upon ClpXP

overexpression, arresting cell growth. In

addition, steady state levels of TrmD were

visualized by western blot (Figure 14, bottom). Cells were first grown in LB media (with trace

amount of glucose) to suppress ClpXP expression. Upon induction with arabinose (final

concentration of 0.2%), the cells were collected and subjected to western blot assay with various

time points. The western blot shows that TrmD levels drop significantly in the cell between 30 –

40 min after ClpXP overexpression.

Having confirmed that colonies fail to grow when TrmD is depleted, and that the

depletion kinetics are fairly rapid, growth curves were obtained to determine the optimal time

after induction for harvesting the cultures for ribosome profiling. Surprisingly, however, the

growth curves of degron and control strains did not show much differences (data not shown). We

hypothesized that the slow decay of residual methylated tRNAs, produced prior to TrmD

Figure 14 TrmD depletion

Cell growths on plate (Top) and western

blot for tagged proteins (Bottom).

52

depletion, left enough tRNA for the cells to continue to function and grow relatively normally

over short time frame.

By diluting the

culture twice back to an

OD600 of 0.1, we found that

differences in growth rates

between the degron and

control strains could be

observed after about 4 h of

induction of the ClpXP

protease (Figure 15).

Without the dilutions, the

cultures reached saturation

at nearly the same rates.

With the dilutions,

however, differences could

be observed when the cultures were still growing in exponential phase. It was important not to

reach saturation because starvation for certain amino acids ensues during stationary phase,

creating a signal of pausing at those codons in the ribosome profiling data that would interfere

with our analyses of the effects of depletion of trmD (Figure 15).

We then monitored the level of m1G37 over the 4 h induction time course. A 14-nt long

primer complementary to tRNALeu/CAG was used for extension by reverse transcriptase (Figure

16, left). Methylation blocks extension, generating a 15 nt product. In the absence of methylation

Figure 15 Growth measurements upon TrmD depletion

Cell growth in LB media. At Time 0, exponentially growing cells

were diluted in fresh media with 0.2% arabinose for ClpXP

induction. Additional dilutions of cell culture media to fresh media

after 1h and 2h.

53

at G37, a 53 nt product is generated. In the degron strain, a gradual depletion m1G37 methylated

tRNA was observed (Figure 16, right). At the same time, newly synthesized tRNA is

unmethylated, as evidenced by the gradual accumulation of the 53 nt product over time in the

degron strain (Figure 16, right). No unmethylated tRNA was evident in the control strain.

3.3.2 Depletion of m1G37 of tRNA Causes Pauses on CCG, CCA, and CGG Codons

Ribosome profiling libraries were generated from cultures of the control and degron

strains after 4 h of induction. We analyzed the data for pauses at specific codons that are read by

tRNAs normally containing m1G37. To detect pauses, we averaged the ribosome occupancy

surrounding instances of the 61 sense codons. Plots of average ribosome occupancy for the

control strain, for example, show very strong pauses when Ser codons are found in the ribosomal

A, P, and E sites (Figure 17). In these plots (Figure 17 and 18), the signal at 0 corresponds to the

Figure 16 Effect of TrmD depletion on tRNA

Primer extension assay using tRNAs as templates for verifying methylation depletion on tRNA.

54

codon is positioned in the P site, in the A site at -3, and in the E site at +3. (Note that the noise

between -10 and -20 corresponds to sequence bias at the 3’-end of the reads and can be ignored).

The Ser codons TCG, TCC, TCT, and TCA have the strong signals in the control strain, shown

in color (Figure 17). Strong pauses on serine codons are a signature of growth in LB media that

is frequently observed in E. coli ribosome profiling. Presumably this occurs because levels of Ser

within the cell are limiting for protein synthesis.

In the degron strain (Figure 18), on the other hand, the usual Ser pauses are not observed

and other pauses have become rate limiting. Ribosome occupancy is strongly enriched at CCG,

CCA, and CGG codons. This result, seemingly representing the effect of depletion of tRNAs

containing m1G37, makes sense because all three of these codons are decoded by tRNAs with

m1G37. Perhaps ribosomes pause at these codons because the concentration of tRNA is reduced

or because unmethylated tRNA behaves poorly in decoding or translocation.

Figure 17 Ribosome occupancy on codons in control strain

High level of ribosome occupancy on some codons (colored) compared to other codons (grey).

55

What is surprising is that only these three codons have strong pauses in the profiling data

from the degron strain, when seven or perhaps nine codons are predicted to be decoded by

tRNAs containing m1G37. Table 1 lists the codons that start with C and are decoded by tRNA

isoacceptors that are candidate substrates of TrmD. Note that the CAN codons encoding His and

Gln are not shown in the table; the corresponding tRNAs contain A37 rather than G37 and are

therefore not substrates of TrmD; pauses are not observed on these codons. The same thing is

true for three of the CGN Arg codons: no pauses are observed for the CGU, CGC, and CGA,

codons that are decoded by a tRNA containing an ICG anticodon with A37 and not G37 and so

not a TrmD substrate. In contrast, the CGG Arg codon whose cognate tRNA has m1G37

modification shows very strong pauses, as noted above. This is perhaps the simplest case.

Figure 18 Ribosome occupancy on codons in degron strain

High level of ribosome occupancy on some codons (colored) compared to other codons

(grey).

56

Surprisingly, the CUN leucine codons do not show pausing, even though they are read by

tRNAs thought to contain m1G37. There is some controversy in the literature over whether E.

coli tRNALeu/GAG (green) is actually methylated; this tRNA decodes the CUC and CUU codons.

Further work will be required to ascertain the methylation status of G37 of this tRNA, but our

data may argue against it. Another possibility is that because both the GAG and CAG anticodon

isoacceptors for tRNALeu have pseudouridine at position 38, they may be able to perform correct

base pairing even when m1G37 is depleted. Only these two tRNAs have pseudouridine at

position 38 among the isoacceptors in Table 1.

The pausing on Pro codons is likewise surprising. All three tRNAPro isoacceptors are

bona fide substrates for TrmD; why are strong pauses observed only at CCG and CCA? One

thought is that the GGG

isoacceptor that reads the

other two codons (CCC and

CCU) can maintain some

activity even in the absence

of methylation. The

possibilities for

frameshifting are increased

dramatically with this

isoacceptor (which has the

four nt anticodon GGGG in

the absence of methylation). Perhaps pausing on the CCC and CCU codons leads to rapid

frameshifting so that strong pauses are not observed.

Codon Anti-codon

Leucine

CUC

GAGm1G37

CUA

CUG CAGm1G37

CUU

Proline

CCU GGGm1G37

UGGm1G37 CCC

CCA

CCG CGGm1G37

Arginine

CGU

ICG

CGC

CGA

CGG CCGm1G37

Table 1 Codons and their anti-codons with modifications

Codons with most ribosome occupancy (red) and isoacceptors with

m1G37 modification.

57

3.4 Materials and Methods

3.4.1 Strains and Construction

The figure 13 shows the scheme of constructing the trmD-deg and trmD-cont cells. First,

the degron- and control-tag followed by tetracycline marker were amplified from plasmids

created by Carr et al, provided by Dr. Sean Moore90.

The primers contain homologous extensions at 5' ends to the trmD-flanking regions

(Forward: CGGAACACGCACAACAGCAACATAAACATGATGGGATGGCGGGTGGCTCC

GACTACAAGG, Reverse: ATAATTTAATCTCTTATCCTGGGTAAACTGATATCTCGGGG

GCTTAGGTCGAGGTGGCCC).

PCR products were electroporated into E. coli recombinogenic strain SM140590 and

homologous recombination was confirmed by colony-PCR using primers targeting 5'-terminal

and 3'-flanking regions of trmD (Forward: ATGTGGATTGGCATAATTAGCCTGTTT

CC, Reverse: GAATTCCGGTTACGAATAGCGATAACCACGCC).

trmD-deg SM1405 and trmD-cont SM1405 were used as a donor for P1 phage lysate

preparation, and the tagged chromosomal trmD loci were transferred by P1 transduction to the

recipient E. coli strain G78, in which the chromosomal clpX gene is knocked out. After

tetracycline marker selection, genotype was confirmed again by colony-PCR.

As described90, trmD-deg G78 strain was then transformed with a library of pClpPX

which contains random mutations at the promoter region of the clpPX genes, and transformant

clones were screened out to pick up the one with highest efficiency of degradation, as proven

also by the rapid degradation at protein level. This specific plasmid clone that confers high

degradation efficiency was extracted from trmD-deg G78 and was transformed into trmD-cont

G78 strain. This pair of trmD-cont G78 and trmD-deg G78 is used for all the experiments.

58

3.4.2 Cell Growth and Preparation of Ribo-seq Libraries

Each cell from degron- and control-strain was grown in LB media supplemented with

0.2% glucose, 0.2% glycerol and an appropriated antibiotic at 37° C for overnight. The next day,

fresh 10-mL LB media supplemented with 0.002% glucose, 5 mM serine and an appropriate

antibiotic were inoculated with overnight cell culture media with 100-fold dilution, followed by

incubation for 1.5-2 h shaking at 37° C.

By measuring the O.D600 to reach up to 0.3-0.4, cell cultures were diluted aiming O.D600

≈ 0.1 in fresh 300 mL LB media supplemented with 0.2% arabinose, 5 mM serine and an

appropriate antibiotic, followed by incubation for 1 h shaking at 37° C. Then, by measuring the

O.D600, cell cultures were further diluted in 300-mL LB media supplemented with 0.2%

arabinose, 5 mM serine and an appropriate antibiotic and this was repeated once more after

another hour. Finally, when the O.D600 reached up to 0.3, cells were harvested by filtering cell

culture using 49 µM filter paper and the cell pellet was flash frozen in liquid nitrogen.

Ribo-seq libraries were generated as described previously66 with some modifications.

Ribosome-protected fragments were size-selected between 10 – 40 nt. Cell lysates were digested

with 3000 U of MNase (Roche) for 1 h at 25 °C.

3.4.3 Data Analysis

Data analysis obtained from Illumina sequencing were described previously66.

59

REFERENCES

1. Ibba, M.; Soll, D., Aminoacyl-tRNAs: setting the limits of the genetic code. Genes Dev

2004, 18 (7), 731-8.

2. Ramakrishnan, V., Ribosome structure and the mechanism of translation. Cell 2002, 108

(4), 557-72.

3. Boelens, R.; Gualerzi, C. O., Structure and function of bacterial initiation factors. Curr

Protein Pept Sci 2002, 3 (1), 107-19.

4. Steitz, J. A.; Jakes, K., How ribosomes select initiator regions in mRNA: base pair

formation between the 3' terminus of 16S rRNA and the mRNA during initiation of protein

synthesis in Escherichia coli. Proc Natl Acad Sci U S A 1975, 72 (12), 4734-8.

5. (a) Shine, J.; Dalgarno, L., The 3'-terminal sequence of Escherichia coli 16S ribosomal

RNA: complementarity to nonsense triplets and ribosome binding sites. Proc Natl Acad Sci U S A

1974, 71 (4), 1342-6; (b) Marzi, S.; Myasnikov, A. G.; Serganov, A.; Ehresmann, C.; Romby, P.;

Yusupov, M.; Klaholz, B. P., Structured mRNAs regulate translation initiation by binding to the

platform of the ribosome. Cell 2007, 130 (6), 1019-31.

6. Shatkin, A. J., Capping of eucaryotic mRNAs. Cell 1976, 9 (4 PT 2), 645-53.

7. Carter, A. P.; Clemons, W. M., Jr.; Brodersen, D. E.; Morgan-Warren, R. J.; Hartsch, T.;

Wimberly, B. T.; Ramakrishnan, V., Crystal structure of an initiation factor bound to the 30S

ribosomal subunit. Science 2001, 291 (5503), 498-501.

8. Ogle, J. M.; Carter, A. P.; Ramakrishnan, V., Insights into the decoding mechanism from

recent ribosome structures. Trends Biochem Sci 2003, 28 (5), 259-266.

9. Diaconu, M.; Kothe, U.; Schlunzen, F.; Fischer, N.; Harms, J. M.; Tonevitsky, A. G.; Stark,

H.; Rodnina, M. V.; Wahl, M. C., Structural basis for the function of the ribosomal L7/12 stalk in

factor binding and GTPase activation. Cell 2005, 121 (7), 991-1004.

10. Berchtold, H.; Reshetnikova, L.; Reiser, C. O.; Schirmer, N. K.; Sprinzl, M.; Hilgenfeld,

R., Crystal structure of active elongation factor Tu reveals major domain rearrangements. Nature

1993, 365 (6442), 126-32.

11. Nissen, P.; Kjeldgaard, M.; Thirup, S.; Clark, B. F.; Nyborg, J., The ternary complex of

aminoacylated tRNA and EF-Tu-GTP. Recognition of a bond and a fold. Biochimie 1996, 78 (11-

12), 921-33.

12. Ogle, J. M.; Brodersen, D. E.; Clemons, W. M.; Tarry, M. J.; Carter, A. P.; Ramakrishnan,

V., Recognition of cognate transfer RNA by the 30S ribosomal subunit. Science 2001, 292 (5518),

897-902.

60

13. Voorhees, R. M.; Schmeing, T. M.; Kelley, A. C.; Ramakrishnan, V., The mechanism for

activation of GTP hydrolysis on the ribosome. Science 2010, 330 (6005), 835-8.

14. Blanchard, S. C.; Gonzalez, R. L.; Kim, H. D.; Chu, S.; Puglisi, J. D., tRNA selection and

kinetic proofreading in translation. Nat Struct Mol Biol 2004, 11 (10), 1008-14.

15. Voorhees, R. M.; Ramakrishnan, V., Structural basis of the translational elongation cycle.

Annu Rev Biochem 2013, 82, 203-36.

16. (a) Li, W.; Frank, J., Transfer RNA in the hybrid P/E state: correlating molecular dynamics

simulations with cryo-EM data. Proc Natl Acad Sci U S A 2007, 104 (42), 16540-5; (b) Flanagan,

J. F. t.; Namy, O.; Brierley, I.; Gilbert, R. J., Direct observation of distinct A/P hybrid-state tRNAs

in translocating ribosomes. Structure 2010, 18 (2), 257-64.

17. Dorner, S.; Brunelle, J. L.; Sharma, D.; Green, R., The hybrid state of tRNA binding is an

authentic translation elongation intermediate. Nat Struct Mol Biol 2006, 13 (3), 234-41.

18. Klaholz, B. P.; Pape, T.; Zavialov, A. V.; Myasnikov, A. G.; Orlova, E. V.; Vestergaard,

B.; Ehrenberg, M.; van Heel, M., Structure of the Escherichia coli ribosomal termination complex

with release factor 2. Nature 2003, 421 (6918), 90-4.

19. Ito, K.; Ebihara, K.; Uno, M.; Nakamura, Y., Conserved motifs in prokaryotic and

eukaryotic polypeptide release factors: tRNA-protein mimicry hypothesis. P Natl Acad Sci USA

1996, 93 (11), 5443-5448.

20. Korostelev, A. A., Structural aspects of translation termination on the ribosome. Rna-a

Publication of the Rna Society 2011, 17 (8), 1409-1421.

21. Mora, L.; Heurgue-Hamard, V.; Champ, S.; Ehrenberg, M.; Kisselev, L. L.; Buckingham,

R. H., The essential role of the invariant GGQ motif in the function and stability in vivo of bacterial

release factors RF1 and RF2. Mol Microbiol 2003, 47 (1), 267-75.

22. Freistroffer, D. V.; Pavlov, M. Y.; MacDougall, J.; Buckingham, R. H.; Ehrenberg, M.,

Release factor RF3 in E.coli accelerates the dissociation of release factors RF1 and RF2 from the

ribosome in a GTP-dependent manner. Embo J 1997, 16 (13), 4126-33.

23. Pavlov, M. Y.; Freistroffer, D. V.; MacDougall, J.; Buckingham, R. H.; Ehrenberg, M.,

Fast recycling of Escherichia coli ribosomes requires both ribosome recycling factor (RRF) and

release factor RF3. Embo J 1997, 16 (13), 4134-41.

24. Doma, M. K.; Parker, R., Endonucleolytic cleavage of eukaryotic mRNAs with stalls in

translation elongation. Nature 2006, 440 (7083), 561-4.

25. Keiler, K. C., Mechanisms of ribosome rescue in bacteria. Nat Rev Microbiol 2015, 13 (5),

285-297.

26. Pech, M.; Nierhaus, K. H., Three mechanisms in Escherichia coli rescue ribosomes stalled

on non-stop mRNAs: one of them requires release factor 2. Mol Microbiol 2012, 86 (1), 6-9.

61

27. Chadani, Y.; Ono, K.; Ozawa, S.; Takahashi, Y.; Takai, K.; Nanamiya, H.; Tozawa, Y.;

Kutsukake, K.; Abo, T., Ribosome rescue by Escherichia coli ArfA (YhdL) in the absence of trans-

translation system. Mol Microbiol 2010, 78 (4), 796-808.

28. Chadani, Y.; Ono, K.; Kutsukake, K.; Abo, T., Escherichia coli YaeJ protein mediates a

novel ribosome-rescue pathway distinct from SsrA- and ArfA-mediated pathways. Mol Microbiol

2011, 80 (3), 772-85.

29. Metzinger, L.; Hallier, M.; Felden, B., The highest affinity binding site of small protein B

on transfer messenger RNA is outside the tRNA domain. RNA 2008, 14 (9), 1761-72.

30. Miller, M. R.; Healey, D. W.; Robison, S. G.; Dewey, J. D.; Buskirk, A. R., The role of

upstream sequences in selecting the reading frame on tmRNA. BMC Biol 2008, 6, 29.

31. Karzai, A. W.; Roche, E. D.; Sauer, R. T., The SsrA-SmpB system for protein tagging,

directed degradation and ribosome rescue. Nat Struct Biol 2000, 7 (6), 449-55.

32. Keiler, K. C., Biology of trans-translation. Annu Rev Microbiol 2008, 62, 133-51.

33. Zechel, D. L.; Konermann, L.; Withers, S. G.; Douglas, D. J., Pre-steady state kinetic

analysis of an enzymatic reaction monitored by time-resolved electrospray ionization mass

spectrometry. Biochemistry-Us 1998, 37 (21), 7664-9.

34. Arava, Y.; Wang, Y.; Storey, J. D.; Liu, C. L.; Brown, P. O.; Herschlag, D., Genome-wide

analysis of mRNA translation profiles in Saccharomyces cerevisiae. Proc Natl Acad Sci U S A

2003, 100 (7), 3889-94.

35. Ingolia, N. T.; Ghaemmaghami, S.; Newman, J. R. S.; Weissman, J. S., Genome-Wide

Analysis in Vivo of Translation with Nucleotide Resolution Using Ribosome Profiling. Science

2009, 324 (5924), 218-223.

36. Ingolia, N. T.; Lareau, L. F.; Weissman, J. S., Ribosome profiling of mouse embryonic

stem cells reveals the complexity and dynamics of mammalian proteomes. Cell 2011, 147 (4), 789-

802.

37. Michel, A. M.; Choudhury, K. R.; Firth, A. E.; Ingolia, N. T.; Atkins, J. F.; Baranov, P. V.,

Observation of dually decoded regions of the human genome using ribosome profiling data.

Genome Res 2012, 22 (11), 2219-2229.

38. Dunn, J. G.; Foo, C. K.; Belletier, N. G.; Gavis, E. R.; Weissman, J. S., Ribosome profiling

reveals pervasive and regulated stop codon readthrough in Drosophila melanogaster. Elife 2013,

2, e01179.

39. Guydosh, N. R.; Green, R., Dom34 rescues ribosomes in 3' untranslated regions. Cell 2014,

156 (5), 950-62.

62

40. Oh, E.; Becker, A. H.; Sandikci, A.; Huber, D.; Chaba, R.; Gloge, F.; Nichols, R. J.; Typas,

A.; Gross, C. A.; Kramer, G.; Weissman, J. S.; Bukau, B., Selective ribosome profiling reveals the

cotranslational chaperone action of trigger factor in vivo. Cell 2011, 147 (6), 1295-308.

41. Li, G. W.; Oh, E.; Weissman, J. S., The anti-Shine-Dalgarno sequence drives translational

pausing and codon choice in bacteria. Nature 2012, 484 (7395), 538-U172.

42. Mohammad, F.; Woolstenhulme, C. J.; Green, R.; Buskirk, A. R., Clarifying the

Translational Pausing Landscape in Bacteria by Ribosome Profiling. Cell Rep 2016, 14 (4), 686-

94.

43. Li, G. W.; Burkhardt, D.; Gross, C.; Weissman, J. S., Quantifying absolute protein

synthesis rates reveals principles underlying allocation of cellular resources. Cell 2014, 157 (3),

624-35.

44. Ogura, T.; Hiraga, S., Mini-F Plasmid Genes That Couple Host-Cell Division to Plasmid

Proliferation. P Natl Acad Sci-Biol 1983, 80 (15), 4784-4788.

45. (a) Bernard, P.; Kezdy, K. E.; Vanmelderen, L.; Steyaert, J.; Wyns, L.; Pato, M. L.; Higgins,

P. N.; Couturier, M., The F-Plasmid Ccdb Protein Induces Efficient Atp-Dependent DNA

Cleavage by Gyrase. J Mol Biol 1993, 234 (3), 534-541; (b) Bernard, P.; Couturier, M., Cell Killing

by the F-Plasmid Ccdb Protein Involves Poisoning of DNA-Topoisomerase-Ii Complexes. J Mol

Biol 1992, 226 (3), 735-745.

46. Tam, J. E.; Kline, B. C., The F-Plasmid Ccd Autorepressor Is a Complex of Ccda and Ccdb

Proteins. Mol Gen Genet 1989, 219 (1-2), 26-32.

47. Jaffe, A.; Ogura, T.; Hiraga, S., Effects of the Ccd Function of the F-Plasmid on Bacterial-

Growth. J Bacteriol 1985, 163 (3), 841-849.

48. Gerdes, K., Toxin-antitoxin modules may regulate synthesis of macromolecules during

nutritional stress. J Bacteriol 2000, 182 (3), 561-572.

49. Lehnherr, H.; Maguin, E.; Jafri, S.; Yarmolinsky, M. B., Plasmid Addiction Genes of

Bacteriophage-P1 - Doc, Which Causes Cell-Death on Curing of Prophage, and Phd, Which

Prevents Host Death When Prophage Is Retained. J Mol Biol 1993, 233 (3), 414-428.

50. (a) Pedersen, K.; Zavialov, A. V.; Pavlov, M. Y.; Elf, J.; Gerdes, K.; Ehrenberg, M., The

bacterial toxin RelE displays codon-specific cleavage of rnRNAs in the ribosomal A site. Cell

2003, 112 (1), 131-140; (b) Christensen, S. K.; Mikkelsen, M.; Pedersen, K.; Gerdes, K., ReIE, a

global inhibitor of translation, is activated during nutritional stress. P Natl Acad Sci USA 2001, 98

(25), 14328-14333.

51. Neubauer, C.; Gao, Y. G.; Andersen, K. R.; Dunham, C. M.; Kelley, A. C.; Hentschel, J.;

Gerdes, K.; Ramakrishnan, V.; Brodersen, D. E., The Structural Basis for mRNA Recognition and

Cleavage by the Ribosome-Dependent Endonuclease ReIE. Cell 2009, 139 (6), 1084-1095.

63

52. Hurley, J. M.; Cruz, J. W.; Ouyang, M.; Woychik, N. A., Bacterial Toxin RelE Mediates

Frequent Codon-independent mRNA Cleavage from the 5 ' End of Coding Regions in Vivo. J Biol

Chem 2011, 286 (17), 14770-14778.

53. Gamper, H. B.; Masuda, I.; Frenkel-Morgenstern, M.; Hou, Y. M., The UGG Isoacceptor

of tRNAPro Is Naturally Prone to Frameshifts. Int J Mol Sci 2015, 16 (7), 14866-83.

54. Brule, H.; Elliott, M.; Redlak, M.; Zehner, Z. E.; Holmes, W. M., Isolation and

characterization of the human tRNA-(N1G37) methyltransferase (TRM5) and comparison to the

Escherichia coli TrmD protein. Biochemistry-Us 2004, 43 (28), 9243-55.

55. Lewis, K., Persister cells, dormancy and infectious disease. Nat Rev Microbiol 2007, 5 (1),

48-56.

56. Veening, J. W.; Smits, W. K.; Kuipers, O. P., Bistability, Epigenetics, and Bet-Hedging in

Bacteria. Annu Rev Microbiol 2008, 62, 193-210.

57. Hauryliuk, V.; Atkinson, G. C.; Murakami, K. S.; Tenson, T.; Gerdes, K., Recent

functional insights into the role of (p)ppGpp in bacterial physiology. Nat Rev Microbiol 2015, 13

(5), 298-309.

58. Christensen, S. K.; Maenhaut-Michel, G.; Mine, N.; Gottesman, S.; Gerdes, K.; Van

Melderen, L., Overproduction of the Lon protease triggers inhibition of translation in Escherichia

coli: involvement of the yefM-yoeB toxin-antitoxin system. Mol Microbiol 2004, 51 (6), 1705-

1717.

59. (a) Gotfredsen, M.; Gerdes, K., The Escherichia coli relBE genes belong to a new toxin-

antitoxin gene family. Mol Microbiol 1998, 29 (4), 1065-76; (b) Christensen, S. K.; Gerdes, K.,

Delayed-relaxed response explained by hyperactivation of RelE. Mol Microbiol 2004, 53 (2), 587-

597.

60. Brzozowska, I.; Zielenkiewicz, U., Regulation of toxin-antitoxin systems by proteolysis.

Plasmid 2013, 70 (1), 33-41.

61. Gerdes, K.; Maisonneuve, E., Bacterial persistence and toxin-antitoxin loci. Annu Rev

Microbiol 2012, 66, 103-23.

62. Maisonneuve, E.; Shakespeare, L. J.; Jorgensen, M. G.; Gerdes, K., Bacterial persistence

by RNA endonucleases. Proc Natl Acad Sci U S A 2011, 108 (32), 13206-11.

63. Griffin, M. A.; Davis, J. H.; Strobel, S. A., Bacterial Toxin RelE: A Highly Efficient

Ribonuclease with Exquisite Substrate Specificity Using Atypical Catalytic Residues.

Biochemistry-Us 2013, 52 (48), 8633-8642.

64. Schifano, J. M.; Vvedenskaya, I. O.; Knoblauch, J. G.; Ouyang, M.; Nickels, B. E.;

Woychik, N. A., An RNA-seq method for defining endoribonuclease cleavage specificity

identifies dual rRNA substrates for toxin MazF-mt3. Nat Commun 2014, 5.

64

65. Christensen, S. K.; Gerdes, K., RelE toxins from bacteria and Archaea cleave mRNAs on

translating ribosomes, which are rescued by tmRNA. Mol Microbiol 2003, 48 (5), 1389-400.

66. Woolstenhulme, C. J.; Guydosh, N. R.; Green, R.; Buskirk, A. R., High-Precision Analysis

of Translational Pausing by Ribosome Profiling in Bacteria Lacking EFP. Cell Rep 2015, 11 (1),

13-21.

67. (a) Nakatogawa, H.; Ito, K., The ribosomal exit tunnel functions as a discriminating gate.

Cell 2002, 108 (5), 629-636; (b) Gong, F.; Yanofsky, C., Instruction of translating ribosome by

nascent peptide. Science 2002, 297 (5588), 1864-1867.

68. Craigen, W. J.; Caskey, C. T., Expression of peptide chain release factor 2 requires high-

efficiency frameshift. Nature 1986, 322 (6076), 273-5.

69. Qian, Q.; Bjork, G. R., Structural requirements for the formation of 1-methylguanosine in

vivo in tRNA(GGG)(Pro) of Salmonella typhimurium. J Mol Biol 1997, 266 (2), 283-296.

70. (a) Holmes, W. M.; Andraos-Selim, C.; Redlak, M., tRNA-m1G methyltransferase

interactions: touching bases with structure. Biochimie 1995, 77 (1-2), 62-5; (b) Bjork, G. R.;

Wikstrom, P. M.; Bystrom, A. S., Prevention of Translational Frameshifting by the Modified

Nucleoside 1-Methylguanosine. Science 1989, 244 (4907), 986-989; (c) Bystrom, A. S.; Bjork, G.

R., Chromosomal location and cloning of the gene (trmD) responsible for the synthesis of tRNA

(m1G) methyltransferase in Escherichia coli K-12. Mol Gen Genet 1982, 188 (3), 440-6.

71. Bystrom, A. S.; Bjork, G. R., The structural gene (trmD) for the

tRNA(m1G)methyltransferase is part of a four polypeptide operon in Escherichia coli K-12. Mol

Gen Genet 1982, 188 (3), 447-54.

72. Christian, T.; Gamper, H.; Hou, Y. M., Conservation of structure and mechanism by Trm5

enzymes. Rna-a Publication of the Rna Society 2013, 19 (9), 1192-1199.

73. Li, Z. W.; Deutscher, M. P., Maturation pathways for E-coli tRNA precursors: A random

multienzyme process in vivo. Cell 1996, 86 (3), 503-512.

74. Frank, D. N.; Pace, N. R., Ribonuclease P: Unity and diversity in a tRNA processing

ribozyme. Annual Review of Biochemistry 1998, 67, 153-180.

75. Li, Z. W.; Deutscher, M. P., RNase E plays an essential role in the maturation of

Escherichia coli tRNA precursors. Rna-a Publication of the Rna Society 2002, 8 (1), 97-109.

76. Perona, J. J.; Rould, M. A.; Steitz, T. A., Structural basis for transfer RNA aminoacylation

by Escherichia coli glutaminyl-tRNA synthetase. Biochemistry-Us 1993, 32 (34), 8758-71.

77. Rould, M. A.; Perona, J. J.; Steitz, T. A., Structural basis of anticodon loop recognition by

glutaminyl-tRNA synthetase. Nature 1991, 352 (6332), 213-8.

65

78. (a) Charette, M.; Gray, M. W., Pseudouridine in RNA: What, where, how, and why. Iubmb

Life 2000, 49 (5), 341-351; (b) Davis, F. F.; Allen, F. W., Ribonucleic acids from yeast which

contain a fifth nucleotide. J Biol Chem 1957, 227 (2), 907-15.

79. Dalluge, J. J.; Hashizume, T.; Sopchik, A. E.; McCloskey, J. A.; Davis, D. R.,

Conformational flexibility in RNA: The role of dihydrouridine. Nucleic Acids Res 1996, 24 (6),

1073-1079.

80. Urbonavicius, J.; Qian, Q.; Durand, J. M.; Hagervall, T. G.; Bjork, G. R., Improvement of

reading frame maintenance is a common function for several tRNA modifications. Embo J 2001,

20 (17), 4863-73.

81. Nasvall, S. J.; Chen, P.; Bjork, G. R., The modified wobble nucleoside uridine-5-oxyacetic

acid in tRNAPro(cmo5UGG) promotes reading of all four proline codons in vivo. RNA 2004, 10

(10), 1662-73.

82. Gamper, H. B.; Masuda, I.; Frenkel-Morgenstern, M.; Hou, Y. M., Maintenance of protein

synthesis reading frame by EF-P and m(1)G37-tRNA. Nat Commun 2015, 6, 7226.

83. Bregeon, D.; Colot, V.; Radman, M.; Taddei, F., Translational misreading: a tRNA

modification counteracts a+2 ribosomal frameshift. Gene Dev 2001, 15 (17), 2295-2306.

84. Christian, T.; Evilia, C.; Williams, S.; Hou, Y. M., Distinct origins of tRNA(m1G37)

methyltransferase. J Mol Biol 2004, 339 (4), 707-19.

85. Ito, T.; Masuda, I.; Yoshida, K.; Goto-Ito, S.; Sekine, S.; Suh, S. W.; Hou, Y. M.;

Yokoyama, S., Structural basis for methyl-donor-dependent and sequence-specific binding to

tRNA substrates by knotted methyltransferase TrmD. Proc Natl Acad Sci U S A 2015, 112 (31),

E4197-205.

86. Elkins, P. A.; Watts, J. M.; Zalacain, M.; van Thiel, A.; Vitazka, P. R.; Redlak, M.;

Andraos-Selim, C.; Rastinejad, F.; Holmes, W. M., Insights into catalysis by a knotted TrmD

tRNA methyltransferase. J Mol Biol 2003, 333 (5), 931-949.

87. O'Dwyer, K.; Watts, J. M.; Biswas, S.; Ambrad, J.; Barber, M.; Brule, H.; Petit, C.; Holmes,

D. J.; Zalacain, M.; Holmes, W. M., Characterization of Streptococcus pneumoniae TrmD, a tRNA

methyltransferase essential for growth. J Bacteriol 2004, 186 (8), 2346-54.

88. Christian, T.; Hou, Y. M., Distinct determinants of tRNA recognition by the TrmD and

Trm5 methyl transferases. J Mol Biol 2007, 373 (3), 623-32.

89. Goto-Ito, S.; Ito, T.; Ishii, R.; Muto, Y.; Bessho, Y.; Yokoyama, S., Crystal structure of

archaeal tRNA(m(1)G37)methyltransferase aTrm5. Proteins 2008, 72 (4), 1274-89.

90. Carr, A. C.; Taylor, K. L.; Osborne, M. S.; Belous, B. T.; Myerson, J. P.; Moore, S. D.,

Rapid depletion of target proteins allows identification of coincident physiological responses. J

Bacteriol 2012, 194 (21), 5932-40.

66

91. Gottesman, S.; Roche, E.; Zhou, Y.; Sauer, R. T., The ClpXP and ClpAP proteases degrade

proteins with carboxy-terminal peptide tails added by the SsrA-tagging system. Genes Dev 1998,

12 (9), 1338-47.